Table of Contents
MySQL Cluster is a high-availability, high-redundancy version of
MySQL adapted for the distributed computing environment. It uses the
NDB Cluster
storage engine to enable running
several MySQL servers in a cluster. This storage engine is available
in MySQL 5.1 binary releases and in RPMs compatible
with most modern Linux distributions.
MySQL 5.1 is currently available and supported on a number of platforms, including Linux, Solaris, Mac OS X, BSD, HP-UX, and many other Unix-style operating systems on a variety of hardware. For exact levels of support available for on specific combinations of operating system versions, operating system distributions, and hardware platforms, please refer to the Cluster Supported Platforms list maintained by the MySQL Support Team on the MySQL AB Web site.
MySQL Cluster is not currently supported on Microsoft Windows. We are working to make Cluster available on all operating systems supported by MySQL, including Windows, and will update the information provided here as this work continues.
This chapter represents a work in progress, and its contents are subject to revision as MySQL Cluster continues to evolve. Additional information regarding MySQL Cluster can be found on the MySQL AB Web site at http://www.mysql.com/products/cluster/.
Additional resources
Answers to some commonly asked questions about Cluster may be found in the Section聽A.10, 鈥淢ySQL 5.1 FAQ 鈥 MySQL Cluster鈥.
The MySQL Cluster mailing list: http://lists.mysql.com/cluster.
The MySQL Cluster Forum: http://forums.mysql.com/list.php?25.
Many MySQL Cluster users and some of the MySQL Cluster developers blog about their experiences with Cluster, and make feeds of these available through PlanetMySQL.
If you are new to MySQL Cluster, you may find our Developer Zone article How to set up a MySQL Cluster for two servers to be helpful.
MySQL Cluster is a technology that enables clustering of in-memory databases in a shared-nothing system. The shared-nothing architecture allows the system to work with very inexpensive hardware, and with a minimum of specific requirements for hardware or software.
MySQL Cluster is designed not to have any single point of failure. For this reason, each component is expected to have its own memory and disk, and the use of shared storage mechanisms such as network shares, network filesystems, and SANs is not recommended or supported.
MySQL Cluster integrates the standard MySQL server with an
in-memory clustered storage engine called NDB
.
In our documentation, the term NDB
refers to
the part of the setup that is specific to the storage engine,
whereas 鈥MySQL Cluster鈥 refers to the combination of
MySQL and the NDB
storage engine.
A MySQL Cluster consists of a set of computers, each running a one or more processes which may include a MySQL server, a data node, a management server, and (possibly) a specialized data access programs. The relationship of these components in a cluster is shown here:
All these programs work together to form a MySQL Cluster. When
data is stored in the NDB Cluster
storage
engine, the tables are stored in the data nodes. Such tables are
directly accessible from all other MySQL servers in the cluster.
Thus, in a payroll application storing data in a cluster, if one
application updates the salary of an employee, all other MySQL
servers that query this data can see this change immediately.
The data stored in the data nodes for MySQL Cluster can be mirrored; the cluster can handle failures of individual data nodes with no other impact than that a small number of transactions are aborted due to losing the transaction state. Because transactional applications are expected to handle transaction failure, this should not be a source of problems.
NDB
is an in-memory
storage engine offering high-availability and data-persistence
features.
The NDB
storage engine can be configured with a
range of failover and load-balancing options, but it is easiest to
start with the storage engine at the cluster level. MySQL
Cluster's NDB
storage engine contains a
complete set of data, dependent only on other data within the
cluster itself.
The cluster portion of MySQL Cluster is currently configured independently of the MySQL servers. In a MySQL Cluster, each part of the cluster is considered to be a node.
Note: In many contexts, the term 鈥node鈥 is used to indicate a computer, but when discussing MySQL Cluster it means a process. It is possible to run any number of nodes on a single computer, for which we use the term cluster host.
(However, it should be noted MySQL does not currently support the use of multiple data nodes on a single computer in a production setting. See Issues exclusive to MySQL Cluster.)
There are three types of cluster nodes, and in a minimal MySQL Cluster configuration, there will be at least three nodes, one of each of these types:
Management node (MGM node): The role of this type of node is to manage the other nodes within the MySQL Cluster, performing such functions as providing configuration data, starting and stopping nodes, running backup, and so forth. Because this node type manages the configuration of the other nodes, a node of this type should be started first, before any other node. An MGM node is started with the command ndb_mgmd.
Data node: This type of node stores cluster data. There are as many data nodes as there are replicas, times the number of fragments. For example, with two replicas, each having two fragments, you will need four data nodes. It is not necessary to have more than one replica. A data node is started with the command ndbd.
SQL node: This is a node that accesses
the cluster data. In the case of MySQL Cluster, an SQL node is
a traditional MySQL server that uses the NDB
Cluster
storage engine. An SQL node is typically
started with the command mysqld
--ndbcluster or by using mysqld
with the ndbcluster
option added to
my.cnf
.
An SQL node is actually just a specialised type of API node, which designates any application which accesses Cluster data. One example of an API node is the ndb_restore utility that is used to restore a cluster backup. It is possible to write such applications using the NDB API.
Important: It is not realistic to expect to employ a three-node setup in a production environment. Such a configuration provides no redundancy; in order to benefit from MySQL Cluster's high-availability features, you must use multiple data and SQL nodes. The use of multiple management nodes is also highly recommended.
For a brief introduction to the relationships between nodes, node groups, replicas, and partitions in MySQL Cluster, see Section聽15.2.1, 鈥淢ySQL Cluster Nodes, Node Groups, Replicas, and Partitions鈥.
Configuration of a cluster involves configuring each individual node in the cluster and setting up individual communication links between nodes. MySQL Cluster is currently designed with the intention that data nodes are homogeneous in terms of processor power, memory space, and bandwidth. In addition, to provide a single point of configuration, all configuration data for the cluster as a whole is located in one configuration file.
The management server (MGM node) manages the cluster configuration file and the cluster log. Each node in the cluster retrieves the configuration data from the management server, and so requires a way to determine where the management server resides. When interesting events occur in the data nodes, the nodes transfer information about these events to the management server, which then writes the information to the cluster log.
In addition, there can be any number of cluster client processes or applications. These are of two types:
Standard MySQL clients: These are no different for MySQL Cluster than they are for standard (non-Cluster) MySQL. In other words, MySQL Cluster can be accessed from existing MySQL applications written in PHP, Perl, C, C++, Java, Python, Ruby, and so on.
Management clients: These clients connect to the management server and provide commands for starting and stopping nodes gracefully, starting and stopping message tracing (debug versions only), showing node versions and status, starting and stopping backups, and so on.
This section discusses the manner in which MySQL Cluster divides and duplicates data for storage.
Central to an understanding of this topic are the following concepts, listed here with brief definitions:
(Data) Node: An ndbd process, which stores a replica 鈥攖hat is, a copy of the partition (see below) assigned to the node group of which the node is a member.
Each data node should be located on a separate computer. While it is also possible to host multiple ndbd processes on a single computer, such a configuration is not supported.
It is common for the terms 鈥node鈥 and 鈥data node鈥 to be used interchangeably when referring to an ndbd process; where mentioned, management (MGM) nodes (ndb_mgmd processes) and SQL nodes (mysqld processes) are specified as such in this discussion.
Node Group: A node group consists of one or more nodes, and stores partitions, or sets of replicas (see next item).
Note: Currently, all node groups in a cluster must have the same number of nodes.
Partition: This is a portion of the data stored by the cluster. There are as many cluster partitions as nodes participating in the cluster. Each node is responsible for keeping at least one copy of any partitions assigned to it (that is, at least one replica) available to the cluster.
A replica belongs entirely to a single node; a node can (and usually does) store several replicas.
Replica: This is a copy of a cluster partition. Each node in a node group stores a replica. Also sometimes known as a partition replica. The number of replicas is equal to the number of nodes per node group.
The following diagram illustrates a MySQL Cluster with four data nodes, arranged in two node groups of two nodes each; nodes 1 and 2 belong to node group 0, and nodes 3 and 4 belong to node group 1. Note that only data (ndbd) nodes are shown here; although a working cluster requires an ndb_mgm process for cluster management and at least one SQL node to access the data stored by the cluster, these have been omitted in the figure for clarity.
The data stored by the cluster is divided into four partitions, numbered 0, 1, 2, and 3. Each partition is stored 鈥 in multiple copies 鈥 on the same node group. Partitions are stored on alternate node groups:
Partition 0 is stored on node group 0; a primary replica (primary copy) is stored on node 1, and a backup replica (backup copy of the partition) is stored on node 2.
Partition 1 is stored on the other node group (node group 1); this partition's primary replica is on node 3, and its backup replica is on node 4.
Partition 2 is stored on node group 0. However, the placing of its two replicas is reversed from that of Partition 0; for Partition 2, the primary replica is stored on node 2, and the backup on node 1.
Partition 3 is stored on node group 1, and the placement of its two replicas are reversed from those of partition 1. That is, its primary replica is located on node 4, with the backup on node 3.
What this means regarding the continued operation of a MySQL Cluster is this: so long as each node group participating in the cluster has at least one node operating, the cluster has a complete copy of all data and remains viable. This is illustrated in the next diagram.
In this example, where the cluster consists of two node groups of two nodes each, any combination of at least one node in node group 0 and at least one node in node group 1 is sufficient to keep the cluster 鈥alive鈥 (indicated by arrows in the diagram). However, if both nodes from either node group fail, the remaining two nodes are not sufficient (shown by the arrows marked out with an X); in either case, the cluster has lost an entire partition and so can no longer provide access to a complete set of all cluster data.
This section is a 鈥How-To鈥 that describes the basics for how to plan, install, configure, and run a MySQL Cluster. Whereas the examples in Section聽15.4, 鈥淢ySQL Cluster Configuration鈥 provide more in-depth information on a variety of clustering options and configuration, the result of following the guidelines and procedures outlined here should be a usable MySQL Cluster which meets the minimum requirements for availability and safeguarding of data.
This section covers hardware and software requirements; networking issues; installation of MySQL Cluster; configuration issues; starting, stopping, and restarting the cluster; loading of a sample database; and performing queries.
Basic Assumptions
This How-To makes the following assumptions:
The cluster setup has four nodes, each on a separate host, and each with a fixed network address on a typical Ethernet as shown here:
Node | IP Address |
Management (MGM) node | 192.168.0.10 |
MySQL server (SQL) node | 192.168.0.20 |
Data (NDBD) node "A" | 192.168.0.30 |
Data (NDBD) node "B" | 192.168.0.40 |
This may be made clearer in the following diagram:
In the interest of simplicity (and reliability), this How-To
uses only numeric IP addresses. However, if DNS resolution is
available on your network, it is possible to use hostnames in
lieu of IP addresses in configuring Cluster. Alternatively,
you can use the /etc/hosts
file or your
operating system's equivalent for providing a means to do host
lookup if such is available.
A common problem when trying to use hostnames for Cluster
nodes arises because of the way in which some operating
systems (including some Linux distributions) set up the
system's own hostname in the /etc/hosts
during installation. Consider two machines with the
hostnames ndb1
and
ndb2
, both in the
cluster
network domain. Red Hat Linux
(including some derivatives such as CentOS and Fedora)
places the following entries in these machines'
/etc/hosts
files:
# ndb1 /etc/hosts
:
127.0.0.1 ndb1.cluster ndb1 localhost.localdomain localhost
# ndb2 /etc/hosts
:
127.0.0.1 ndb2.cluster ndb2 localhost.localdomain localhost
SuSE Linux (including OpenSuSE) places these entries in the
machines' /etc/hosts
files:
# ndb1 /etc/hosts
:
127.0.0.1 localhost
127.0.0.2 ndb1.cluster ndb1
# ndb2 /etc/hosts
:
127.0.0.1 localhost
127.0.0.2 ndb2.cluster ndb2
In both instances, ndb1
routes
ndb1.cluster
to a loopback IP address,
but gets a public IP address from DNS for
ndb2.cluster
, while
ndb2
routes
ndb2.cluster
to a loopback address and
obtains a public address for
ndb1.cluster
. The result is that each
data node connects to the management server, but cannot tell
when any other data nodes have connected, and so the data
nodes appear to hang while starting.
You should also be aware that you cannot mix
localhost
and other hostnames or IP
addresses in config.ini
. For these
reasons, the solution in such cases (other than to use IP
addresses for all
config.ini
HostName
entries) is to remove the fully qualified hostnames from
/etc/hosts
and use these in
config.ini
for all cluster hosts.
Each host in our scenario is an Intel-based desktop PC running a common, generic Linux distribution installed to disk in a standard configuration, and running no unnecessary services. The core OS with standard TCP/IP networking capabilities should be sufficient. Also for the sake of simplicity, we also assume that the filesystems on all hosts are set up identically. In the event that they are not, you will need to adapt these instructions accordingly.
Standard 100 Mbps or 1 gigabit Ethernet cards are installed on each machine, along with the proper drivers for the cards, and that all four hosts are connected via a standard-issue Ethernet networking appliance such as a switch. (All machines should use network cards with the same throughout. That is, all four machines in the cluster should have 100 Mbps cards or all four machines should have 1 Gbps cards.) MySQL Cluster will work in a 100 Mbps network; however, gigabit Ethernet will provide better performance.
Note that MySQL Cluster is not intended for use in a network for which throughput is less than 100 Mbps. For this reason (among others), attempting to run a MySQL Cluster over a public network such as the Internet is not likely to be successful, and is not recommended.
For our sample data, we will use the world
database which is available for download from the MySQL AB Web
site. As this database takes up a relatively small amount of
space, we assume that each machine has 256MB RAM, which should
be sufficient for running the operating system, host NDB
process, and (for the data nodes) for storing the database.
Although we refer to a Linux operating system in this How-To, the instructions and procedures that we provide here should be easily adaptable to other supported operating systems. We also assume that you already know how to perform a minimal installation and configuration of the operating system with networking capability, or that you are able to obtain assistance in this elsewhere if needed.
We discuss MySQL Cluster hardware, software, and networking requirements in somewhat greater detail in the next section. (See Section聽15.3.1, 鈥淗ardware, Software, and Networking鈥.)
One of the strengths of MySQL Cluster is that it can be run on commodity hardware and has no unusual requirements in this regard, other than for large amounts of RAM, due to the fact that all live data storage is done in memory. (Note that this is not the case with Disk Data tables 鈥 see Section聽15.11, 鈥淢ySQL Cluster Disk Data Tables鈥, for more information about these.) Naturally, multiple and faster CPUs will enhance performance. Memory requirements for other Cluster processes are relatively small.
The software requirements for Cluster are also modest. Host operating systems do not require any unusual modules, services, applications, or configuration to support MySQL Cluster. For supported operating systems, a standard installation should be sufficient. The MySQL software requirements are simple: all that is needed is a production release of MySQL 5.1 to have Cluster support. It is not necessary to compile MySQL yourself merely to be able to use Cluster. In this How-To, we assume that you are using the server binary appropriate to your operating system, available via the MySQL software downloads page at http://dev.mysql.com/downloads/.
For inter-node communication, Cluster supports TCP/IP networking in any standard topology, and the minimum expected for each host is a standard 100 Mbps Ethernet card, plus a switch, hub, or router to provide network connectivity for the cluster as a whole. We strongly recommend that a MySQL Cluster be run on its own subnet which is not shared with non-Cluster machines for the following reasons:
Security: Communications between Cluster nodes are not encrypted or shielded in any way. The only means of protecting transmissions within a MySQL Cluster is to run your Cluster on a protected network. If you intend to use MySQL Cluster for Web applications, the cluster should definitely reside behind your firewall and not in your network's De-Militarized Zone (DMZ) or elsewhere.
Efficiency: Setting up a MySQL Cluster on a private or protected network allows the cluster to make exclusive use of bandwidth between cluster hosts. Using a separate switch for your MySQL Cluster not only helps protect against unauthorized access to Cluster data, it also ensures that Cluster nodes are shielded from interference caused by transmissions between other computers on the network. For enhanced reliability, you can use dual switches and dual cards to remove the network as a single point of failure; many device drivers support failover for such communication links.
It is also possible to use the high-speed Scalable Coherent Interface (SCI) with MySQL Cluster, but this is not a requirement. See Section聽15.12, 鈥淯sing High-Speed Interconnects with MySQL Cluster鈥, for more about this protocol and its use with MySQL Cluster.
Each MySQL Cluster host computer running data or SQL nodes must have installed on it a MySQL server binary. For management nodes, it is not necessary to install the MySQL server binary, but you do have to install the MGM server daemon (ndb_mgmd). It is also a good idea to install the management client (ndb_mgm) on the management server host. This section covers the steps necessary to install the correct binaries for each type of Cluster node.
MySQL AB provides precompiled binaries that support Cluster, and
there is generally no need to compile these yourself. Therefore,
the first step in the installation process for each cluster host
is to download the file
mysql-5.1.18-beta-pc-linux-gnu-i686.tar.gz
from the MySQL downloads
area. We assume that you have placed it in each
machine's /var/tmp
directory. (If you do
require a custom binary, see
Section聽2.9.3, 鈥淚nstalling from the Development Source Tree鈥.)
RPMs are also available for both 32-bit and 64-bit Linux platforms. For a MySQL Cluster, three RPMs are required:
The Server RPM (for
example,
MySQL-server-5.1.18-beta-0.glibc23.i386.rpm
),
which supplies the core files needed to run a MySQL Server
and a MySQL Server binary with clustering support.
The NDB Cluster - Storage
engine RPM (for example,
MySQL-ndb-storage-5.1.18-beta-0.glibc23.i386.rpm
),
which supplies the MySQL Cluster data node binary
(ndbd).
The NDB Cluster - Storage engine
management RPM (for example,
MySQL-ndb-management-5.1.18-beta-0.glibc23.i386.rpm
),
which provides the MySQL Cluster management server binary
(ndb_mgmd).
In addition, you should also obtain the
NDB Cluster - Storage engine basic
tools RPM (for example,
MySQL-ndb-tools-5.1.18-beta-0.glibc23.i386.rpm
),
which supplies several useful applications for working with a
MySQL Cluster. The most important of the these is the MySQL
Cluster management client (ndb_mgm). The
NDB Cluster - Storage engine extra
tools RPM (for example,
MySQL-ndb-extra-5.1.18-beta-0.glibc23.i386.rpm
)
contains some additional testing and monitoring programs, but is
not required to install a MySQL Cluster. (For more information
about these additional programs, see
Section聽15.9, 鈥淐luster Utility Programs鈥.)
The MySQL version number in the RPM filenames (shown here as
5.1.18-beta
) can vary according to the
version which you are actually using. It is very
important that all of the Cluster RPMs to be installed have the
same MySQL version number. The
glibc
version number (if present 鈥
shown here as glibc23
), and architecture
designation (shown here as i386
) should be
appropriate to the machine on which the RPM is to be installed.
See Section聽2.4, 鈥淚nstalling MySQL on Linux鈥, for general information about installing MySQL using RPMs supplied by MySQL AB.
After installing from RPM, you still need to configure the cluster as discussed in Section聽15.3.3, 鈥淢ulti-Computer Configuration鈥.
Note: After completing the installation, do not yet start any of the binaries. We show you how to do so following the configuration of all nodes.
Data and SQL Node Installation 鈥
.tar.gz
Binary
On each of the machines designated to host data or SQL nodes,
perform the following steps as the system
root
user:
Check your /etc/passwd
and
/etc/group
files (or use whatever tools
are provided by your operating system for managing users and
groups) to see whether there is already a
mysql
group and mysql
user on the system. Some OS distributions create these as
part of the operating system installation process. If they
are not already present, create a new
mysql
user group, and then add a
mysql
user to this group:
shell>groupadd mysql
shell>useradd -g mysql mysql
The syntax for useradd and groupadd may differ slightly on different versions of Unix, or they may have different names such as adduser and addgroup.
Change location to the directory containing the downloaded
file, unpack the archive, and create a symlink to the
mysql
directory. Note that the actual
file and directory names will vary according to the MySQL
version number.
shell>cd /var/tmp
shell>tar -C /usr/local -xzvf mysql-5.1.18-beta-pc-linux-gnu-i686.tar.gz
shell>ln -s /usr/local/mysql-5.1.18-beta-pc-linux-gnu-i686 /usr/local/mysql
Change location to the mysql
directory
and run the supplied script for creating the system
databases:
shell>cd mysql
shell>scripts/mysql_install_db --user=mysql
Set the necessary permissions for the MySQL server and data directories:
shell>chown -R root .
shell>chown -R mysql data
shell>chgrp -R mysql .
Note that the data directory on each machine hosting a data
node is /usr/local/mysql/data
. We will
use this piece of information when we configure the
management node. (See
Section聽15.3.3, 鈥淢ulti-Computer Configuration鈥.)
Copy the MySQL startup script to the appropriate directory, make it executable, and set it to start when the operating system is booted up:
shell>cp support-files/mysql.server /etc/rc.d/init.d/
shell>chmod +x /etc/rc.d/init.d/mysql.server
shell>chkconfig --add mysql.server
(The startup scripts directory may vary depending on your
operating system and version 鈥 for example, in some
Linux distributions, it is
/etc/init.d
.)
Here we use Red Hat's chkconfig for creating links to the startup scripts; use whatever means is appropriate for this purpose on your operating system and distribution, such as update-rc.d on Debian.
Remember that the preceding steps must be performed separately on each machine where a data node or SQL node is to reside.
SQL Node Installation 鈥 RPM Files
On each machine to be used for hosting a cluster SQL node, install the MySQL RPM by executing the following command as the system root user, replacing the name shown for the RPM as necessary to match the name of the RPM downloaded from the MySQL AB web site:
shell> rpm -Uhv MySQL-server-5.1.18-beta-0.glibc23.i386.rpm
This installs the MySQL server binary
(mysqld) in the
/usr/sbin
directory, as well as all needed
MySQL Server support files. It also installs the
mysql.server and
mysqld_safe startup scripts in
/usr/share/mysql
and
/usr/bin
, respectively. The RPM installer
should take care of general configuration issues (such as
creating the mysql
user and group, if needed)
automatically.
Data Node Installation 鈥 RPM Files
On a computer that is to host a cluster data node it is necessary to install only the NDB Cluster - Storage engine RPM. To do so, copy this RPM to the data node host, and run the following command as the system root user, replacing the name shown for the RPM as necessary to match that of the RPM downloaded from the MySQL AB web site:
shell> rpm -Uhv MySQL-ndb-storage-5.1.18-beta-0.glibc23.i386.rpm
The previous command installs the MySQL Cluster data node binary
(ndbd) in the /usr/sbin
directory.
Management Node Installation 鈥
.tar.gz
Binary
Installation for the management (MGM) node does not require
installation of the mysqld binary. Only the
binaries for the MGM server and client are required, which can
be found in the downloaded archive. Again, we assume that you
have placed this file in /var/tmp
.
As system root
(that is, after using
sudo, su root, or your
system's equivalent for temporarily assuming the system
administrator account's privileges), perform the following steps
to install ndb_mgmd and
ndb_mgm on the Cluster management node host:
Change location to the /var/tmp
directory, and extract the ndb_mgm and
ndb_mgmd from the archive into a suitable
directory such as /usr/local/bin
:
shell>cd /var/tmp
shell>tar -zxvf mysql-5.1.18-beta-pc-linux-gnu-i686.tar.gz
shell>cd mysql-5.1.18-beta-pc-linux-gnu-i686
shell>cp /bin/ndb_mgm* /usr/local/bin
(You can safely delete the directory created by unpacking
the downloaded archive, and the files it contains, from
/var/tmp
once
ndb_mgm and ndb_mgmd
have been copied to the executables directory.)
Change location to the directory into which you copied the files, and then make both of them executable:
shell>cd /usr/local/bin
shell>chmod +x ndb_mgm*
Management Node Installation 鈥 RPM File
To install the MySQL Cluster management server, it is necessary only to use the NDB Cluster - Storage engine management RPM. Copy this RPM to the computer intended to host the management node, and then install it by running the following command as the system root user (replace the name shown for the RPM as necessary to match that of the Storage engine management RPM downloaded from the MySQL AB web site):
shell> rpm -Uhv MySQL-ndb-management-5.1.18-beta-0.glibc23.i386.rpm
This installs the management server binary
(ndb_mgmd) to the
/usr/sbin
directory.
You should also install the NDB
management
client, which is supplied by the Storage
engine basic tools RPM. Copy this RPM to the same
computer as the management node, and then install it by running
the following command as the system root user (again, replace
the name shown for the RPM as necessary to match that of the
Storage engine basic tools RPM
downloaded from the MySQL AB web site):
shell> rpm -Uhv MySQL-ndb-tools-5.1.18-beta-0.glibc23.i386.rpm
The Storage engine basic tools
RPM installs the MySQL Cluster management client
(ndb_mgm) to the
/usr/bin
directory.
In Section聽15.3.3, 鈥淢ulti-Computer Configuration鈥, we create configuration files for all of the nodes in our example Cluster.
For our four-node, four-host MySQL Cluster, it is necessary to write four configuration files, one per node host.
Each data node or SQL node requires a
my.cnf
file that provides two pieces of
information: a connectstring that
tells the node where to find the MGM node, and a line
telling the MySQL server on this host (the machine hosting
the data node) to run in NDB mode.
For more information on connectstrings, see Section聽15.4.4.2, 鈥淭he Cluster Connectstring鈥.
The management node needs a config.ini
file telling it how many replicas to maintain, how much
memory to allocate for data and indexes on each data node,
where to find the data nodes, where to save data to disk on
each data node, and where to find any SQL nodes.
Configuring the Storage and SQL Nodes
The my.cnf
file needed for the data nodes
is fairly simple. The configuration file should be located in
the /etc
directory and can be edited using
any text editor. (Create the file if it does not exist.) For
example:
shell> vi /etc/my.cnf
We show vi being used here to create the file, but any text editor should work just as well.
For each data node and SQL node in our example setup,
my.cnf
should look like this:
# Options for mysqld process: [MYSQLD] ndbcluster # run NDB storage engine ndb-connectstring=192.168.0.10 # location of management server # Options for ndbd process: [MYSQL_CLUSTER] ndb-connectstring=192.168.0.10 # location of management server
After entering the preceding information, save this file and exit the text editor. Do this for the machines hosting data node 鈥A鈥, data node 鈥B鈥, and the SQL node.
Important: Once you have
started a mysqld process with the
ndbcluster
and
ndb-connectstring
parameters in the
[MYSQLD]
section of the
my.cnf
file as shown previously, you cannot
execute any CREATE TABLE
or ALTER
TABLE
statements without having actually started the
cluster. Otherwise, these statements will fail with an error.
This is by design.
Configuring the Management Node
The first step in configuring the MGM node is to create the
directory in which the configuration file can be found and then
to create the file itself. For example (running as
root
):
shell>mkdir /var/lib/mysql-cluster
shell>cd /var/lib/mysql-cluster
shell>vi config.ini
For our representative setup, the
config.ini
file should read as follows:
# Options affecting ndbd processes on all data nodes: [NDBD DEFAULT] NoOfReplicas=2 # Number of replicas DataMemory=80M # How much memory to allocate for data storage IndexMemory=18M # How much memory to allocate for index storage # For DataMemory and IndexMemory, we have used the # default values. Since the "world" database takes up # only about 500KB, this should be more than enough for # this example Cluster setup. # TCP/IP options: [TCP DEFAULT] portnumber=2202 # This the default; however, you can use any # port that is free for all the hosts in the cluster # Note: It is recommended that you do not specify the # portnumber at all and allow the default value to be # used instead # Management process options: [NDB_MGMD] hostname=192.168.0.10 # Hostname or IP address of MGM node datadir=/var/lib/mysql-cluster # Directory for MGM node log files # Options for data node "A": [NDBD] # (one [NDBD] section per data node) hostname=192.168.0.30 # Hostname or IP address datadir=/usr/local/mysql/data # Directory for this data node's data files # Options for data node "B": [NDBD] hostname=192.168.0.40 # Hostname or IP address datadir=/usr/local/mysql/data # Directory for this data node's data files # SQL node options: [MYSQLD] hostname=192.168.0.20 # Hostname or IP address # (additional mysqld connections can be # specified for this node for various # purposes such as running ndb_restore)
(Note: The
world
database can be downloaded from
http://dev.mysql.com/doc/, where it can be found listed
under 鈥Examples鈥.)
After all the configuration files have been created and these minimal options have been specified, you are ready to proceed with starting the cluster and verifying that all processes are running. We discuss how this is done in Section聽15.3.4, 鈥淚nitial Startup鈥.
For more detailed information about the available MySQL Cluster configuration parameters and their uses, see Section聽15.4.4, 鈥淐onfiguration File鈥, and Section聽15.4, 鈥淢ySQL Cluster Configuration鈥. For configuration of MySQL Cluster as relates to making backups, see Section聽15.8.4, 鈥淐onfiguration for Cluster Backup鈥.
Note: The default port for Cluster management nodes is 1186; the default port for data nodes is 2202. Beginning with MySQL 5.0.3, this restriction is lifted, and the cluster automatically allocates ports for data nodes from those that are already free.
Starting the cluster is not very difficult after it has been configured. Each cluster node process must be started separately, and on the host where it resides. The management node should be started first, followed by the data nodes, and then finally by any SQL nodes:
On the management host, issue the following command from the system shell to start the management node process:
shell> ndb_mgmd -f /var/lib/mysql-cluster/config.ini
Note that ndb_mgmd must be told where to
find its configuration file, using the -f
or --config-file
option. (See
Section聽15.6.3, 鈥ndb_mgmd 鈥 The Management Server Process鈥, for
details.)
On each of the data node hosts, run this command to start the ndbd process for the first time:
shell> ndbd --initial
Note that it is very important to use the
--initial
parameter
only when starting
ndbd for the first time, or when
restarting after a backup/restore operation or a
configuration change. This is because the
--initial
option causes the node to delete
any files created by earlier ndbd
instances that are needed for recovery, including the
recovery log files.
An exception to this is that --initial
does
not delete Disk Data files. If you do need to perform an
initial restart of the cluster, you must delete any existing
Disk Data log files and data files manually.
If you used RPM files to install MySQL on the cluster host where the SQL node is to reside, you can (and should) use the supplied startup script to start the MySQL server process on the SQL node.
If all has gone well, and the cluster has been set up correctly, the cluster should now be operational. You can test this by invoking the ndb_mgm management node client. The output should look like that shown here, although you might see some slight differences in the output depending upon the exact version of MySQL that you are using:
shell>ndb_mgm
-- NDB Cluster -- Management Client -- ndb_mgm>SHOW
Connected to Management Server at: localhost:1186 Cluster Configuration --------------------- [ndbd(NDB)] 2 node(s) id=2 @192.168.0.30 (Version: 5.1.18-beta, Nodegroup: 0, Master) id=3 @192.168.0.40 (Version: 5.1.18-beta, Nodegroup: 0) [ndb_mgmd(MGM)] 1 node(s) id=1 @192.168.0.10 (Version: 5.1.18-beta) [mysqld(SQL)] 1 node(s) id=4 (Version: 5.1.18-beta)
Note: The SQL node is
referenced here as [mysqld(API)]
. This is
perfectly normal, and reflects the fact that the
mysqld process is acting as a cluster API
node.
You should now be ready to work with databases, tables, and data in MySQL Cluster. See Section聽15.3.5, 鈥淟oading Sample Data and Performing Queries鈥, for a brief discussion.
Working with data in MySQL Cluster is not much different from doing so in MySQL without Cluster. There are two points to keep in mind:
For a table to be replicated in the cluster, it must use the
NDB Cluster
storage engine. To specify
this, use the ENGINE=NDB
or
ENGINE=NDBCLUSTER
table option. You can add
this option when creating the table:
CREATE TABLEtbl_name
(column_definitions
) ENGINE=NDBCLUSTER;
Alternatively, for an existing table that uses a different
storage engine, use ALTER TABLE
to change
the table to use NDB Cluster
:
ALTER TABLE tbl_name
ENGINE=NDBCLUSTER;
Each NDB
table must
have a primary key. If no primary key is defined by the user
when a table is created, the NDB Cluster
storage engine automatically generates a hidden one.
(Note: This hidden key
takes up space just as does any other table index. It is not
uncommon to encounter problems due to insufficient memory
for accommodating these automatically created indexes.)
If you are importing tables from an existing database using the
output of mysqldump, you can open the SQL
script in a text editor and add the ENGINE
option to any table creation statements, or replace any existing
ENGINE
(or TYPE
) options.
Suppose that you have the world
sample
database on another MySQL server that does not support MySQL
Cluster, and you want to export the City
table:
shell> mysqldump --add-drop-table world City > city_table.sql
The resulting city_table.sql
file will
contain this table creation statement (and the
INSERT
statements necessary to import the
table data):
DROP TABLE IF EXISTS `City`;
CREATE TABLE `City` (
`ID` int(11) NOT NULL auto_increment,
`Name` char(35) NOT NULL default '',
`CountryCode` char(3) NOT NULL default '',
`District` char(20) NOT NULL default '',
`Population` int(11) NOT NULL default '0',
PRIMARY KEY (`ID`)
) ENGINE=MyISAM DEFAULT CHARSET=latin1;
INSERT INTO `City` VALUES (1,'Kabul','AFG','Kabol',1780000);
INSERT INTO `City` VALUES (2,'Qandahar','AFG','Qandahar',237500);
INSERT INTO `City` VALUES (3,'Herat','AFG','Herat',186800);
(remaining INSERT statements omitted)
You need to make sure that MySQL uses the NDB
storage engine for this table. There are two ways that this can
be accomplished. One of these is to modify the table definition
before importing it into the Cluster
database. Using the City
table as an example,
modify the ENGINE
option of the definition as
follows:
DROP TABLE IF EXISTS `City`;
CREATE TABLE `City` (
`ID` int(11) NOT NULL auto_increment,
`Name` char(35) NOT NULL default '',
`CountryCode` char(3) NOT NULL default '',
`District` char(20) NOT NULL default '',
`Population` int(11) NOT NULL default '0',
PRIMARY KEY (`ID`)
) ENGINE=NDBCLUSTER DEFAULT CHARSET=latin1;
INSERT INTO `City` VALUES (1,'Kabul','AFG','Kabol',1780000);
INSERT INTO `City` VALUES (2,'Qandahar','AFG','Qandahar',237500);
INSERT INTO `City` VALUES (3,'Herat','AFG','Herat',186800);
(remaining INSERT statements omitted)
This must be done for the definition of each table that is to be
part of the clustered database. The easiest way to accomplish
this is to do a search-and-replace on the file that contains the
definitions and replace all instances of
TYPE=
or
engine_name
ENGINE=
with engine_name
ENGINE=NDBCLUSTER
. If you do not want to
modify the file, you can use the unmodified file to create the
tables, and then use ALTER TABLE
to change
their storage engine. The particulars are given later in this
section.
Assuming that you have already created a database named
world
on the SQL node of the cluster, you can
then use the mysql command-line client to
read city_table.sql
, and create and
populate the corresponding table in the usual manner:
shell> mysql world < city_table.sql
It is very important to keep in mind that the preceding command
must be executed on the host where the SQL node is running (in
this case, on the machine with the IP address
192.168.0.20
).
To create a copy of the entire world
database
on the SQL node, use mysqldump on the
non-cluster server to export the database to a file named
world.sql
; for example, in the
/tmp
directory. Then modify the table
definitions as just described and import the file into the SQL
node of the cluster like this:
shell> mysql world < /tmp/world.sql
If you save the file to a different location, adjust the preceding instructions accordingly.
It is important to note that NDB Cluster
in
MySQL 5.1 does not support autodiscovery of
databases. (See Section聽15.13, 鈥淜nown Limitations of MySQL Cluster鈥.)
This means that, once the world
database and
its tables have been created on one data node, you need to issue
the CREATE SCHEMA world
statement on all
other SQL nodes attached to the cluster. However, once this has
been done, all of the SQL nodes can 鈥see鈥 the
tables without any further action being required.
Running SELECT
queries on the SQL node is no
different from running them on any other instance of a MySQL
server. To run queries from the command line, you first need to
log in to the MySQL Monitor in the usual way (specify the
root
password at the Enter
password:
prompt):
shell> mysql -u root -p
Enter password:
Welcome to the MySQL monitor. Commands end with ; or \g.
Your MySQL connection id is 1 to server version: 5.1.18-beta
Type 'help;' or '\h' for help. Type '\c' to clear the buffer.
mysql>
We simply use the MySQL server's root
account
and assume that you have followed the standard security
precautions for installing a MySQL server, including setting a
strong root
password. For more information,
see Section聽2.10.3, 鈥淪ecuring the Initial MySQL Accounts鈥.
It is worth taking into account that Cluster nodes do
not make use of the MySQL privilege system
when accessing one another. Setting or changing MySQL user
accounts (including the root
account) effects
only applications that access the SQL node, not interaction
between nodes.
If you did not modify the ENGINE
clauses in
the table definitions prior to importing the SQL script, you
should run the following statements at this point:
mysql>USE world;
mysql>ALTER TABLE City ENGINE=NDBCLUSTER;
mysql>ALTER TABLE Country ENGINE=NDBCLUSTER;
mysql>ALTER TABLE CountryLanguage ENGINE=NDBCLUSTER;
Selecting a database and running a SELECT query against a table in that database is also accomplished in the usual manner, as is exiting the MySQL Monitor:
mysql>USE world;
mysql>SELECT Name, Population FROM City ORDER BY Population DESC LIMIT 5;
+-----------+------------+ | Name | Population | +-----------+------------+ | Bombay | 10500000 | | Seoul | 9981619 | | S茫o Paulo | 9968485 | | Shanghai | 9696300 | | Jakarta | 9604900 | +-----------+------------+ 5 rows in set (0.34 sec) mysql>\q
Bye shell>
Applications that use MySQL can employ standard APIs to access
NDB
tables. It is important to remember that
your application must access the SQL node, and not the
management or data nodes. This brief example shows how we might
execute the SELECT
statement just shown by
using the PHP 5.X mysqli
extension running on
a Web server elsewhere on the network:
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
"http://www.w3.org/TR/html4/loose.dtd">
<html>
<head>
<meta http-equiv="Content-Type"
content="text/html; charset=iso-8859-1">
<title>SIMPLE mysqli SELECT</title>
</head>
<body>
<?php
# connect to SQL node:
$link = new mysqli('192.168.0.20', 'root', 'root_password
', 'world');
# parameters for mysqli constructor are:
# host, user, password, database
if( mysqli_connect_errno() )
die("Connect failed: " . mysqli_connect_error());
$query = "SELECT Name, Population
FROM City
ORDER BY Population DESC
LIMIT 5";
# if no errors...
if( $result = $link->query($query) )
{
?>
<table border="1" width="40%" cellpadding="4" cellspacing ="1">
<tbody>
<tr>
<th width="10%">City</th>
<th>Population</th>
</tr>
<?
# then display the results...
while($row = $result->fetch_object())
printf(<tr>\n <td align=\"center\">%s</td><td>%d</td>\n</tr>\n",
$row->Name, $row->Population);
?>
</tbody
</table>
<?
# ...and verify the number of rows that were retrieved
printf("<p>Affected rows: %d</p>\n", $link->affected_rows);
}
else
# otherwise, tell us what went wrong
echo mysqli_error();
# free the result set and the mysqli connection object
$result->close();
$link->close();
?>
</body>
</html>
We assume that the process running on the Web server can reach the IP address of the SQL node.
In a similar fashion, you can use the MySQL C API, Perl-DBI, Python-mysql, or MySQL AB's own Connectors to perform the tasks of data definition and manipulation just as you would normally with MySQL.
To shut down the cluster, enter the following command in a shell on the machine hosting the MGM node:
shell> ndb_mgm -e shutdown
The -e
option here is used to pass a command to
the ndb_mgm client from the shell. See
Section聽4.3.1, 鈥淯sing Options on the Command Line鈥. The command causes the
ndb_mgm, ndb_mgmd, and any
ndbd processes to terminate gracefully. Any
SQL nodes can be terminated using mysqladmin
shutdown and other means.
To restart the cluster, run these commands:
On the management host (192.168.0.10
in
our example setup):
shell> ndb_mgmd -f /var/lib/mysql-cluster/config.ini
On each of the data node hosts
(192.168.0.30
and
192.168.0.40
):
shell> ndbd
Remember not to invoke this command
with the --initial
option when restarting
an NDBD node normally.
On the SQL host (192.168.0.20
):
shell> mysqld_safe &
For information on making Cluster backups, see Section聽15.8.2, 鈥淯sing The Management Client to Create a Backup鈥.
To restore the cluster from backup requires the use of the ndb_restore command. This is covered in Section聽15.8.3, 鈥ndb_restore 鈥 Restore a Cluster Backup鈥.
More information on configuring MySQL Cluster can be found in Section聽15.4, 鈥淢ySQL Cluster Configuration鈥.
A MySQL server that is part of a MySQL Cluster differs in only one
respect from a normal (non-clustered) MySQL server, in that it
employs the NDB Cluster
storage engine. This
engine is also referred to simply as NDB
, and
the two forms of the name are synonymous.
To avoid unnecessary allocation of resources, the server is
configured by default with the NDB
storage
engine disabled. To enable NDB
, you must modify
the server's my.cnf
configuration file, or
start the server with the --ndbcluster
option.
The MySQL server is a part of the cluster, so it also must know
how to access an MGM node to obtain the cluster configuration
data. The default behavior is to look for the MGM node on
localhost
. However, should you need to specify
that its location is elsewhere, this can be done in
my.cnf
or on the MySQL server command line.
Before the NDB
storage engine can be used, at
least one MGM node must be operational, as well as any desired
data nodes.
NDB
, the Cluster storage engine, is available
in binary distributions for Linux, Mac OS X, and Solaris. We are
working to make Cluster run on all operating systems supported
by MySQL, including Windows.
If you choose to build from a source tarball or the MySQL
5.1 BitKeeper tree, be sure to use the
--with-ndbcluster
option when running
configure. You can also use the
BUILD/compile-pentium-max build script. This
script includes OpenSSL, so you must either have or obtain
OpenSSL to build successfully, or else modify
compile-pentium-max to exclude this
requirement. Of course, you can also just follow the standard
instructions for compiling your own binaries, and then perform
the usual tests and installation procedure. See
Section聽2.9.3, 鈥淚nstalling from the Development Source Tree鈥.
You should also note that compile-pentium-max
installs MySQL to the directory
/usr/local/mysql
, placing all MySQL Cluster
executables, scripts, databases, and support files in
subdirectories under this directory. If this is not what you
desire, be sure to modify the script accordingly.
In the next few sections, we assume that you are already familiar with installing MySQL, and here we cover only the differences between configuring MySQL Cluster and configuring MySQL without clustering. (See Chapter聽2, Installing and Upgrading MySQL, if you require more information about the latter.)
You will find Cluster configuration easiest if you have already
have all management and data nodes running first; this is likely
to be the most time-consuming part of the configuration. Editing
the my.cnf
file is fairly straightforward,
and this section will cover only any differences from
configuring MySQL without clustering.
To familiarize you with the basics, we will describe the simplest possible configuration for a functional MySQL Cluster. After this, you should be able to design your desired setup from the information provided in the other relevant sections of this chapter.
First, you need to create a configuration directory such as
/var/lib/mysql-cluster
, by executing the
following command as the system root
user:
shell> mkdir /var/lib/mysql-cluster
In this directory, create a file named
config.ini
that contains the following
information. Substitute appropriate values for
HostName
and DataDir
as
necessary for your system.
# file "config.ini" - showing minimal setup consisting of 1 data node, # 1 management server, and 3 MySQL servers. # The empty default sections are not required, and are shown only for # the sake of completeness. # Data nodes must provide a hostname but MySQL Servers are not required # to do so. # If you don't know the hostname for your machine, use localhost. # The DataDir parameter also has a default value, but it is recommended to # set it explicitly. # Note: DB, API, and MGM are aliases for NDBD, MYSQLD, and NDB_MGMD # respectively. DB and API are deprecated and should not be used in new # installations. [NDBD DEFAULT] NoOfReplicas= 1 [MYSQLD DEFAULT] [NDB_MGMD DEFAULT] [TCP DEFAULT] [NDB_MGMD] HostName= myhost.example.com [NDBD] HostName= myhost.example.com DataDir= /var/lib/mysql-cluster [MYSQLD] [MYSQLD] [MYSQLD]
You can now start the ndb_mgmd management
server. By default, it attempts to read the
config.ini
file in its current working
directory, so change location into the directory where the file
is located and then invoke ndb_mgmd:
shell>cd /var/lib/mysql-cluster
shell>ndb_mgmd
Then start a single data node by running
ndbd. When starting ndbd
for a given data node for the very first time, you should use
the --initial
option as shown here:
shell> ndbd --initial
For subsequent ndbd starts, you will
generally want to omit the
--initial
option:
shell> ndbd
The reason for omitting --initial
on subsequent
restarts is that this option causes ndbd to
delete and re-create all existing data and log files (as well as
all table metadata) for this data node. One exception to this
rule about not using --initial
except for the
first ndbd invocation is that you use it when
restarting the cluster and restoring from backup after adding
new data nodes.
By default, ndbd looks for the management
server at localhost
on port 1186.
Note: If you have installed
MySQL from a binary tarball, you will need to specify the path
of the ndb_mgmd and ndbd
servers explicitly. (Normally, these will be found in
/usr/local/mysql/bin
.)
Finally, change location to the MySQL data directory (usually
/var/lib/mysql
or
/usr/local/mysql/data
), and make sure that
the my.cnf
file contains the option
necessary to enable the NDB storage engine:
[mysqld] ndbcluster
You can now start the MySQL server as usual:
shell> mysqld_safe --user=mysql &
Wait a moment to make sure the MySQL server is running properly.
If you see the notice mysql ended
, check the
server's .err
file to find out what went
wrong.
If all has gone well so far, you now can start using the
cluster. Connect to the server and verify that the
NDBCLUSTER
storage engine is enabled:
shell>mysql
Welcome to the MySQL monitor. Commands end with ; or \g. Your MySQL connection id is 1 to server version: 5.1.18-beta Type 'help;' or '\h' for help. Type '\c' to clear the buffer. mysql>SHOW ENGINES\G
... *************************** 12. row *************************** Engine: NDBCLUSTER Support: YES Comment: Clustered, fault-tolerant, memory-based tables *************************** 13. row *************************** Engine: NDB Support: YES Comment: Alias for NDBCLUSTER ...
The row numbers shown in the preceding example output may be different from those shown on your system, depending upon how your server is configured.
Try to create an NDBCLUSTER
table:
shell>mysql
mysql>USE test;
Database changed mysql>CREATE TABLE ctest (i INT) ENGINE=NDBCLUSTER;
Query OK, 0 rows affected (0.09 sec) mysql>SHOW CREATE TABLE ctest \G
*************************** 1. row *************************** Table: ctest Create Table: CREATE TABLE `ctest` ( `i` int(11) default NULL ) ENGINE=ndbcluster DEFAULT CHARSET=latin1 1 row in set (0.00 sec)
To check that your nodes were set up properly, start the management client:
shell> ndb_mgm
Use the SHOW command from within the management client to obtain a report on the cluster's status:
NDB> SHOW
Cluster Configuration
---------------------
[ndbd(NDB)] 1 node(s)
id=2 @127.0.0.1 (Version: 3.5.3, Nodegroup: 0, Master)
[ndb_mgmd(MGM)] 1 node(s)
id=1 @127.0.0.1 (Version: 3.5.3)
[mysqld(API)] 3 node(s)
id=3 @127.0.0.1 (Version: 3.5.3)
id=4 (not connected, accepting connect from any host)
id=5 (not connected, accepting connect from any host)
At this point, you have successfully set up a working MySQL
Cluster. You can now store data in the cluster by using any
table created with ENGINE=NDBCLUSTER
or its
alias ENGINE=NDB
.
Configuring MySQL Cluster requires working with two files:
my.cnf
: Specifies options for all MySQL
Cluster executables. This file, with which you should be
familiar with from previous work with MySQL, must be
accessible by each executable running in the cluster.
config.ini
: This file is read only by
the MySQL Cluster management server, which then distributes
the information contained therein to all processes
participating in the cluster.
config.ini
contains a description of
each node involved in the cluster. This includes
configuration parameters for data nodes and configuration
parameters for connections between all nodes in the cluster.
For a quick reference to the sections that can appear in
this file, and what sorts of configuration parameters may be
placed in each section, see
Sections
of the config.ini
File.
We are continuously making improvements in Cluster configuration and attempting to simplify this process. Although we strive to maintain backward compatibility, there may be times when introduce an incompatible change. In such cases we will try to let Cluster users know in advance if a change is not backward compatible. If you find such a change and we have not documented it, please report it in the MySQL bugs database using the instructions given in Section聽1.8, 鈥淗ow to Report Bugs or Problems鈥.
To support MySQL Cluster, you will need to update
my.cnf
as shown in the following example.
Note that the options shown here should not be confused with
those that are used in config.ini
files.
You may also specify these parameters on the command line when
invoking the executables.
# my.cnf # example additions to my.cnf for MySQL Cluster # (valid in MySQL 5.1) # enable ndbcluster storage engine, and provide connectstring for # management server host (default port is 1186) [mysqld] ndbcluster ndb-connectstring=ndb_mgmd.mysql.com # provide connectstring for management server host (default port: 1186) [ndbd] connect-string=ndb_mgmd.mysql.com # provide connectstring for management server host (default port: 1186) [ndb_mgm] connect-string=ndb_mgmd.mysql.com # provide location of cluster configuration file [ndb_mgmd] config-file=/etc/config.ini
(For more information on connectstrings, see Section聽15.4.4.2, 鈥淭he Cluster Connectstring鈥.)
# my.cnf # example additions to my.cnf for MySQL Cluster # (will work on all versions) # enable ndbcluster storage engine, and provide connectstring for management # server host to the default port 1186 [mysqld] ndbcluster ndb-connectstring=ndb_mgmd.mysql.com:1186
Important: Once you have
started a mysqld process with the
ndbcluster
and
ndb-connectstring
parameters in the
[MYSQLD]
in the my.cnf
file as shown previously, you cannot execute any
CREATE TABLE
or ALTER
TABLE
statements without having actually started the
cluster. Otherwise, these statements will fail with an error.
This is by design.
You may also use a separate [mysql_cluster]
section in the cluster my.cnf
file for
settings to be read and used by all executables:
# cluster-specific settings [mysql_cluster] ndb-connectstring=ndb_mgmd.mysql.com:1186
For additional NDB
variables that can be
set in the my.cnf
file, see
Section聽5.2.3, 鈥淪ystem Variables鈥.
The configuration file is named
config.ini
by default. It is read by
ndb_mgmd at startup and can be placed
anywhere. Its location and name are specified by using
--config-file=
on the ndb_mgmd command line. If the
configuration file is not specified,
ndb_mgmd by default tries to read a file
named path_name
config.ini
located in the current
working directory.
Currently, the configuration file is in INI format, which
consists of sections preceded by section headings (surrounded
by square brackets), followed by the appropriate parameter
names and values. One deviation from the standard INI format
is that the parameter name and value can be separated by a
colon (鈥:
鈥) as well as the
equals sign (鈥=
鈥). Another
deviation is that sections are not uniquely identified by
section name. Instead, unique sections (such as two different
nodes of the same type) are identified by a unique ID
specified as a parameter within the section.
Default values are defined for most parameters, and can also
be specified in config.ini
. To create a
default value section, simply add the word
DEFAULT
to the section name. For example,
an [NDBD]
section contains parameters that
apply to a particular data node, whereas an [NDBD
DEFAULT]
section contains parameters that apply to
all data nodes. Suppose that all data nodes should use the
same data memory size. To configure them all, create an
[NDBD DEFAULT]
section that contains a
DataMemory
line to specify the data memory
size.
At a minimum, the configuration file must define the computers and nodes involved in the cluster and on which computers these nodes are located. An example of a simple configuration file for a cluster consisting of one management server, two data nodes and two MySQL servers is shown here:
# file "config.ini" - 2 data nodes and 2 SQL nodes # This file is placed in the startup directory of ndb_mgmd (the # management server) # The first MySQL Server can be started from any host. The second # can be started only on the host mysqld_5.mysql.com [NDBD DEFAULT] NoOfReplicas= 2 DataDir= /var/lib/mysql-cluster [NDB_MGMD] Hostname= ndb_mgmd.mysql.com DataDir= /var/lib/mysql-cluster [NDBD] HostName= ndbd_2.mysql.com [NDBD] HostName= ndbd_3.mysql.com [MYSQLD] [MYSQLD] HostName= mysqld_5.mysql.com
Note that each node has its own section in the
config.ini
. For instance, this cluster
has two data nodes, so the preceding configuration file
contains two [NDBD]
sections defining these
nodes.
Sections of the
config.ini
File
There are six different sections that you can use in the
config.ini
configuration file, as
described in the following list:
[COMPUTER]
: Defines cluster hosts. This
is not required to configure a viable MySQL Cluster, but
be may used as a convenience when setting up a large
cluster. See
Section聽15.4.4.3, 鈥淒efining Cluster Computers鈥, for
more information.
[NDBD]
: Defines a cluster data node
(ndbd process). See
Section聽15.4.4.5, 鈥淒efining Data Nodes鈥, for
details.
[MYSQLD]
: Defines the cluster's MySQL
server nodes (also called SQL or API nodes). For a
discussion of SQL node configuration, see
Section聽15.4.4.6, 鈥淒efining SQL and Other API Nodes鈥.
[MGM]
or [NDB_MGMD]
:
Defines a cluster management server (MGM) node. For
information concerning the configuration of MGM nodes, see
Section聽15.4.4.4, 鈥淒efining the Management Server鈥.
[TCP]
: Defines a TCP/IP connection
between cluster nodes, with TCP/IP being the default
connection protocol. Normally, [TCP]
or
[TCP DEFAULT]
sections are not required
to set up a MySQL Cluster, as the cluster handles this
automatically; however, it may be necessary in some
situations to override the defaults provided by the
cluster. See
Section聽15.4.4.7, 鈥淐luster TCP/IP Connections鈥, for
information about available TCP/IP configuration
parameters and how to use them. (You may also find
Section聽15.4.4.8, 鈥淭CP/IP Connections Using Direct Connections鈥 to
be of interest in some cases.)
[SHM]
: Defines shared-memory
connections between nodes. In MySQL 5.1, it
is enabled by default, but should still be considered
experimental. For a discussion of SHM interconnects, see
Section聽15.4.4.9, 鈥淪hared-Memory Connections鈥.
[SCI]
:Defines Scalable
Coherent Interface connections between cluster
data nodes. Such connections require software which, while
freely available, is not part of the MySQL Cluster
distribution, as well as specialised hardware. See
Section聽15.4.4.10, 鈥淪CI Transport Connections鈥 for
detailed information about SCI interconnects.
You can define DEFAULT
values for each
section. All Cluster parameter names are case-insensitive,
which differs from parameters specified in
my.cnf
or my.ini
files.
With the exception of the MySQL Cluster management server (ndb_mgmd), each node that is part of a MySQL Cluster requires a connectstring that points to the management server's location. This connectstring is used in establishing a connection to the management server as well as in performing other tasks depending on the node's role in the cluster. The syntax for a connectstring is as follows:
<connectstring> := [<nodeid-specification>,]<host-specification>[,<host-specification>] <nodeid-specification> :=node_id
<host-specification> :=host_name
[:port_num
]
node_id
is an integer larger than 1 which
identifies a node in config.ini
.
host_name
is a string representing
a valid Internet host name or IP address.
port_num
is an integer referring to
a TCP/IP port number.
example 1 (long): "nodeid=2,myhost1:1100,myhost2:1100,192.168.0.3:1200" example 2 (short): "myhost1"
All nodes will use localhost:1186
as the
default connectstring value if none is provided. If
port_num
is omitted from the
connectstring, the default port is 1186. This port should
always be available on the network because it has been
assigned by IANA for this purpose (see
http://www.iana.org/assignments/port-numbers
for details).
By listing multiple
<host-specification>
values, it is
possible to designate several redundant management servers. A
cluster node will attempt to contact successive management
servers on each host in the order specified, until a
successful connection has been established.
There are a number of different ways to specify the connectstring:
Each executable has its own command-line option which enables specifying the management server at startup. (See the documentation for the respective executable.)
It is also possible to set the connectstring for all nodes
in the cluster at once by placing it in a
[mysql_cluster]
section in the
management server's my.cnf
file.
For backward compatibility, two other options are available, using the same syntax:
Set the NDB_CONNECTSTRING
environment
variable to contain the connectstring.
Write the connectstring for each executable into a
text file named Ndb.cfg
and place
this file in the executable's startup directory.
However, these are now deprecated and should not be used for new installations.
The recommended method for specifying the connectstring is to
set it on the command line or in the
my.cnf
file for each executable.
The maximum length of a connectstring is 1024 characters.
The [COMPUTER]
section has no real
significance other than serving as a way to avoid the need of
defining host names for each node in the system. All
parameters mentioned here are required.
The [NDB_MGMD]
section is used to configure
the behavior of the management server.
[MGM]
can be used as an alias; the two
section names are equivalent. All parameters in the following
list are optional and assume their default values if omitted.
Note: If neither the
ExecuteOnComputer
nor the
HostName
parameter is present, the default
value localhost
will be assumed for both.
Each node in the cluster has a unique identity, which is represented by an integer value in the range 1 to 63 inclusive. This ID is used by all internal cluster messages for addressing the node.
This refers to the Id
set for one of
the computers defined in a [COMPUTER]
section of the config.ini
file.
This is the port number on which the management server listens for configuration requests and management commands.
Specifying this parameter defines the hostname of the
computer on which the management node is to reside. To
specify a hostname other than
localhost
, either this parameter or
ExecuteOnComputer
is required.
This parameter specifies where to send cluster logging
information. There are three options in this regard
鈥 CONSOLE
,
SYSLOG
, and FILE
鈥 with FILE
being the default:
CONSOLE
outputs the log to
stdout
:
CONSOLE
SYSLOG
sends the log to a
syslog
facility, possible values
being one of auth
,
authpriv
, cron
,
daemon
, ftp
,
kern
, lpr
,
mail
, news
,
syslog
, user
,
uucp
, local0
,
local1
, local2
,
local3
, local4
,
local5
, local6
,
or local7
.
Note: Not every facility is necessarily supported by every operating system.
SYSLOG:facility=syslog
FILE
pipes the cluster log output
to a regular file on the same machine. The following
values can be specified:
filename
: The name of the log
file.
maxsize
: The maximum size (in
bytes) to which the file can grow before logging
rolls over to a new file. When this occurs, the
old log file is renamed by appending
.N
to the filename,
where N
is the next
number not yet used with this name.
maxfiles
: The maximum number of
log files.
FILE:filename=cluster.log,maxsize=1000000,maxfiles=6
The default value for the FILE
parameter is
FILE:filename=ndb_
,
where node_id
_cluster.log,maxsize=1000000,maxfiles=6node_id
is the ID of
the node.
It is possible to specify multiple log destinations separated by semicolons as shown here:
CONSOLE;SYSLOG:facility=local0;FILE:filename=/var/log/mgmd
This parameter is used to define which nodes can act as
arbitrators. Only management nodes and SQL nodes can be
arbitrators. ArbitrationRank
can take
one of the following values:
0
: The node will never be used as
an arbitrator.
1
: The node has high priority; that
is, it will be preferred as an arbitrator over
low-priority nodes.
2
: Indicates a low-priority node
which be used as an arbitrator only if a node with a
higher priority is not available for that purpose.
Normally, the management server should be configured as an
arbitrator by setting its
ArbitrationRank
to 1 (the default
value) and that of all SQL nodes to 0.
An integer value which causes the management server's responses to arbitration requests to be delayed by that number of milliseconds. By default, this value is 0; it is normally not necessary to change it.
This specifies the directory where output files from the
management server will be placed. These files include
cluster log files, process output files, and the daemon's
process ID (PID) file. (For log files, this location can
be overridden by setting the FILE
parameter for LogDestination
as
discussed previously in this section.)
The default value for this parameter is the directory in which ndb_mgmd is located.
The [NDBD]
and [NDBD DEFAULT] sections are
used to configure the behavior of the cluster's data nodes.
There are many parameters which control buffer sizes, pool
sizes, timeouts, and so forth. The only mandatory parameters
are:
Either ExecuteOnComputer
or
HostName
, which must be defined in the
local [NDBD]
section.
The parameter NoOfReplicas
, which must
be defined in the [NDBD DEFAULT] section, as it is common
to all Cluster data nodes.
Most data node parameters are set in the [NDBD
DEFAULT]
section. Only those parameters explicitly
stated as being able to set local values are allowed to be
changed in the [NDBD]
section. Where
present, HostName
, Id
and ExecuteOnComputer
must be defined in the local
[NDBD]
section, and not in any other
section of config.ini
. In other words,
settings for these parameters are specific to one data node.
For those parameters affecting memory usage or buffer sizes,
it is possible to use K
,
M
, or G
as a suffix to
indicate units of 1024, 1024脳1024, or
1024脳1024脳1024. (For example,
100K
means 100 脳 1024 = 102400.)
Parameter names and values are currently case-sensitive.
The Id
value (that is, the data node
identifier) can be allocated on the command line when the node
is started or in the configuration file.
This is the node ID used as the address of the node for all cluster internal messages. This is an integer in the range 1 to 63 inclusive. Each node in the cluster must have a unique identity.
This refers to the Id
set for one of
the computers defined in a [COMPUTER]
section.
Specifying this parameter defines the hostname of the
computer on which the data node is to reside. To specify a
hostname other than localhost
, either
this parameter or ExecuteOnComputer
is
required.
Each node in the cluster uses a port to connect to other nodes. This port is used also for non-TCP transporters in the connection setup phase. The default port is allocated dynamically in such a way as to ensure that no two nodes on the same computer receive the same port number, so it should not normally be necessary to specify a value for this parameter.
This global parameter can be set only in the
[NDBD DEFAULT]
section, and defines the
number of replicas for each table stored in the cluster.
This parameter also specifies the size of node groups. A
node group is a set of nodes all storing the same
information.
Node groups are formed implicitly. The first node group is
formed by the set of data nodes with the lowest node IDs,
the next node group by the set of the next lowest node
identities, and so on. By way of example, assume that we
have 4 data nodes and that NoOfReplicas
is set to 2. The four data nodes have node IDs 2, 3, 4 and
5. Then the first node group is formed from nodes 2 and 3,
and the second node group by nodes 4 and 5. It is
important to configure the cluster in such a manner that
nodes in the same node groups are not placed on the same
computer because a single hardware failure would cause the
entire cluster to crash.
If no node IDs are provided, the order of the data nodes
will be the determining factor for the node group. Whether
or not explicit assignments are made, they can be viewed
in the output of the management client's
SHOW
statement.
There is no default value for
NoOfReplicas
; the maximum possible
value is 4.
Important: The value for
this parameter must divide evenly into the number of data
nodes in the cluster. For example, if there are two data
nodes, then NoOfReplicas
must be equal
to either 1 or 2, since 2/3 and 2/4 both yield fractional
values; if there are four data nodes, then
NoOfReplicas
must be equal to 1, 2, or
4.
This parameter specifies the directory where trace files, log files, pid files and error logs are placed.
This parameter specifies the directory where all files
created for metadata, REDO logs, UNDO logs (for Disk Data
tables) and data files are placed. The default is the
directory specified by DataDir
.
Note: This directory must
exist before the ndbd process is
initiated.
The recommended directory hierarchy for MySQL Cluster
includes /var/lib/mysql-cluster
,
under which a directory for the node's filesystem is
created. The name of this subdirectory contains the node
ID. For example, if the node ID is 2, this subdirectory is
named ndb_2_fs
.
This parameter specifies the directory in which backups
are placed. If omitted, the default backup location is the
directory named BACKUP
under the
location specified by the
FileSystemPath
parameter. (See above.)
Data Memory, Index Memory, and String Memory
DataMemory
and
IndexMemory
are [NDBD]
parameters specifying the size of memory segments used to
store the actual records and their indexes. In setting values
for these, it is important to understand how
DataMemory
and
IndexMemory
are used, as they usually need
to be updated to reflect actual usage by the cluster:
This parameter defines the amount of space (in bytes) available for storing database records. The entire amount specified by this value is allocated in memory, so it is extremely important that the machine has sufficient physical memory to accommodate it.
The memory allocated by DataMemory
is
used to store both the actual records and indexes. There
is a 16-byte overhead on each record; an additional amount
for each record is incurred because it is stored in a 32KB
page with 128 byte page overhead (see below). There is
also a small amount wasted per page due to the fact that
each record is stored in only one page.
For variable-size table attributes in MySQL 5.1, the data
is stored on separate datapages, allocated from
DataMemory
. Variable-length records use
a fixed-size part with an extra overhead of 4 bytes to
reference the variable-size part. The variable-size part
has 2 bytes overhead plus 2 bytes per attribute.
The maximum record size is currently 8052 bytes.
The memory space defined by DataMemory
is also used to store ordered indexes, which use about 10
bytes per record. Each table row is represented in the
ordered index. A common error among users is to assume
that all indexes are stored in the memory allocated by
IndexMemory
, but this is not the case:
Only primary key and unique hash indexes use this memory;
ordered indexes use the memory allocated by
DataMemory
. However, creating a primary
key or unique hash index also creates an ordered index on
the same keys, unless you specify USING
HASH
in the index creation statement. This can
be verified by running ndb_desc -d
db_name
table_name
in the
management client.
The memory space allocated by
DataMemory
consists of 32KB pages,
which are allocated to table fragments. Each table is
normally partitioned into the same number of fragments as
there are data nodes in the cluster. Thus, for each node,
there are the same number of fragments as are set in
NoOfReplicas
.
Once a page has been allocated, it is currently not
possible to return it to the pool of free pages, except by
deleting the table. (This also means that
DataMemory
pages, once allocated to a
given table, cannot be used by other tables.) Performing a
node recovery also compresses the partition because all
records are inserted into empty partitions from other live
nodes.
The DataMemory
memory space also
contains UNDO information: For each update, a copy of the
unaltered record is allocated in the
DataMemory
. There is also a reference
to each copy in the ordered table indexes. Unique hash
indexes are updated only when the unique index columns are
updated, in which case a new entry in the index table is
inserted and the old entry is deleted upon commit. For
this reason, it is also necessary to allocate enough
memory to handle the largest transactions performed by
applications using the cluster. In any case, performing a
few large transactions holds no advantage over using many
smaller ones, for the following reasons:
Large transactions are not any faster than smaller ones
Large transactions increase the number of operations that are lost and must be repeated in event of transaction failure
Large transactions use more memory
The default value for DataMemory
is
80MB; the minimum is 1MB. There is no maximum size, but in
reality the maximum size has to be adapted so that the
process does not start swapping when the limit is reached.
This limit is determined by the amount of physical RAM
available on the machine and by the amount of memory that
the operating system may commit to any one process. 32-bit
operating systems are generally limited to 2鈥4GB per
process; 64-bit operating systems can use more. For large
databases, it may be preferable to use a 64-bit operating
system for this reason.
This parameter controls the amount of storage used for hash indexes in MySQL Cluster. Hash indexes are always used for primary key indexes, unique indexes, and unique constraints. Note that when defining a primary key and a unique index, two indexes will be created, one of which is a hash index used for all tuple accesses as well as lock handling. It is also used to enforce unique constraints.
The size of the hash index is 25 bytes per record, plus the size of the primary key. For primary keys larger than 32 bytes another 8 bytes is added.
The default value for IndexMemory
is
18MB. The minimum is 1MB.
This parameter determines how much memory is allocated for
strings such as table names, and is specified in an
[NDBD]
or [NDBD
DEFAULT]
section of the
config.ini
file. A value between
0
and 100
inclusive
is interpreted as a percent of the maxmimum default value,
which is calculated based on a number of factors including
the number of tables, maximum table name size, maximum
size of .FRM
files,
MaxNoOfTriggers
, maximum column name
size, and maximum default column value. In general it is
safe to assume that the maximum default value is
approximately 5 MB for a MySQL Cluster having 1000 tables.
A value greater than 100
is interpreted
as a number of bytes.
The default value is 5
鈥 that is,
5 percent of the default maximum, or roughly 5 KB. (Note
that this is a change from previous versions of MySQL
Cluster.)
Under most circumstances, the default value should be
sufficient, but when you have a great many Cluster tables
(1000 or more), it is possible to get Error 773
Out of string memory, please modify
StringMemory config parameter: Permanent error: Schema
error, in which case you should increase this
value. 25
(25 percent) is not
excessive, and should prevent this error from recurring in
all but the most extreme conditions.
The following example illustrates how memory is used for a table. Consider this table definition:
CREATE TABLE example ( a INT NOT NULL, b INT NOT NULL, c INT NOT NULL, PRIMARY KEY(a), UNIQUE(b) ) ENGINE=NDBCLUSTER;
For each record, there are 12 bytes of data plus 12 bytes
overhead. Having no nullable columns saves 4 bytes of
overhead. In addition, we have two ordered indexes on columns
a
and b
consuming
roughly 10 bytes each per record. There is a primary key hash
index on the base table using roughly 29 bytes per record. The
unique constraint is implemented by a separate table with
b
as primary key and a
as a column. This other table consumes an additional 29 bytes
of index memory per record in the example
table as well 8 bytes of record data plus 12 bytes of
overhead.
Thus, for one million records, we need 58MB for index memory to handle the hash indexes for the primary key and the unique constraint. We also need 64MB for the records of the base table and the unique index table, plus the two ordered index tables.
You can see that hash indexes takes up a fair amount of memory space; however, they provide very fast access to the data in return. They are also used in MySQL Cluster to handle uniqueness constraints.
Currently, the only partitioning algorithm is hashing and ordered indexes are local to each node. Thus, ordered indexes cannot be used to handle uniqueness constraints in the general case.
An important point for both IndexMemory
and
DataMemory
is that the total database size
is the sum of all data memory and all index memory for each
node group. Each node group is used to store replicated
information, so if there are four nodes with two replicas,
there will be two node groups. Thus, the total data memory
available is 2 脳 DataMemory
for each
data node.
It is highly recommended that DataMemory
and IndexMemory
be set to the same values
for all nodes. Data distribution is even over all nodes in the
cluster, so the maximum amount of space available for any node
can be no greater than that of the smallest node in the
cluster.
DataMemory
and
IndexMemory
can be changed, but decreasing
either of these can be risky; doing so can easily lead to a
node or even an entire MySQL Cluster that is unable to restart
due to there being insufficient memory space. Increasing these
values should be acceptable, but it is recommended that such
upgrades are performed in the same manner as a software
upgrade, beginning with an update of the configuration file,
and then restarting the management server followed by
restarting each data node in turn.
Updates do not increase the amount of index memory used. Inserts take effect immediately; however, rows are not actually deleted until the transaction is committed.
The next three [NDBD]
parameters that we
discuss are important because they affect the number of
parallel transactions and the sizes of transactions that can
be handled by the system.
MaxNoOfConcurrentTransactions
sets the
number of parallel transactions possible in a node.
MaxNoOfConcurrentOperations
sets the number
of records that can be in update phase or locked
simultaneously.
Both of these parameters (especially
MaxNoOfConcurrentOperations
) are likely
targets for users setting specific values and not using the
default value. The default value is set for systems using
small transactions, to ensure that these do not use excessive
memory.
For each active transaction in the cluster there must be a record in one of the cluster nodes. The task of coordinating transactions is spread among the nodes. The total number of transaction records in the cluster is the number of transactions in any given node times the number of nodes in the cluster.
Transaction records are allocated to individual MySQL servers. Normally, there is at least one transaction record allocated per connection that using any table in the cluster. For this reason, one should ensure that there are more transaction records in the cluster than there are concurrent connections to all MySQL servers in the cluster.
This parameter must be set to the same value for all cluster nodes.
Changing this parameter is never safe and doing so can cause a cluster to crash. When a node crashes, one of the nodes (actually the oldest surviving node) will build up the transaction state of all transactions ongoing in the crashed node at the time of the crash. It is thus important that this node has as many transaction records as the failed node.
The default value is 4096.
It is a good idea to adjust the value of this parameter according to the size and number of transactions. When performing transactions of only a few operations each and not involving a great many records, there is no need to set this parameter very high. When performing large transactions involving many records need to set this parameter higher.
Records are kept for each transaction updating cluster data, both in the transaction coordinator and in the nodes where the actual updates are performed. These records contain state information needed to find UNDO records for rollback, lock queues, and other purposes.
This parameter should be set to the number of records to be updated simultaneously in transactions, divided by the number of cluster data nodes. For example, in a cluster which has four data nodes and which is expected to handle 1,000,000 concurrent updates using transactions, you should set this value to 1000000 / 4 = 250000.
Read queries which set locks also cause operation records to be created. Some extra space is allocated within individual nodes to accommodate cases where the distribution is not perfect over the nodes.
When queries make use of the unique hash index, there are actually two operation records used per record in the transaction. The first record represents the read in the index table and the second handles the operation on the base table.
The default value is 32768.
This parameter actually handles two values that can be configured separately. The first of these specifies how many operation records are to be placed with the transaction coordinator. The second part specifies how many operation records are to be local to the database.
A very large transaction performed on an eight-node
cluster requires as many operation records in the
transaction coordinator as there are reads, updates, and
deletes involved in the transaction. However, the
operation records of the are spread over all eight nodes.
Thus, if it is necessary to configure the system for one
very large transaction, it is a good idea to configure the
two parts separately.
MaxNoOfConcurrentOperations
will always
be used to calculate the number of operation records in
the transaction coordinator portion of the node.
It is also important to have an idea of the memory requirements for operation records. These consume about 1KB per record.
By default, this parameter is calculated as 1.1 脳
MaxNoOfConcurrentOperations
. This fits
systems with many simultaneous transactions, none of them
being very large. If there is a need to handle one very
large transaction at a time and there are many nodes, it
is a good idea to override the default value by explicitly
specifying this parameter.
The next set of [NDBD]
parameters is used
to determine temporary storage when executing a statement that
is part of a Cluster transaction. All records are released
when the statement is completed and the cluster is waiting for
the commit or rollback.
The default values for these parameters are adequate for most situations. However, users with a need to support transactions involving large numbers of rows or operations may need to increase these values to enable better parallelism in the system, whereas users whose applications require relatively small transactions can decrease the values to save memory.
MaxNoOfConcurrentIndexOperations
For queries using a unique hash index, another temporary
set of operation records is used during a query's
execution phase. This parameter sets the size of that pool
of records. Thus, this record is allocated only while
executing a part of a query. As soon as this part has been
executed, the record is released. The state needed to
handle aborts and commits is handled by the normal
operation records, where the pool size is set by the
parameter MaxNoOfConcurrentOperations
.
The default value of this parameter is 8192. Only in rare cases of extremely high parallelism using unique hash indexes should it be necessary to increase this value. Using a smaller value is possible and can save memory if the DBA is certain that a high degree of parallelism is not required for the cluster.
The default value of
MaxNoOfFiredTriggers
is 4000, which is
sufficient for most situations. In some cases it can even
be decreased if the DBA feels certain the need for
parallelism in the cluster is not high.
A record is created when an operation is performed that affects a unique hash index. Inserting or deleting a record in a table with unique hash indexes or updating a column that is part of a unique hash index fires an insert or a delete in the index table. The resulting record is used to represent this index table operation while waiting for the original operation that fired it to complete. This operation is short-lived but can still require a large number of records in its pool for situations with many parallel write operations on a base table containing a set of unique hash indexes.
The memory affected by this parameter is used for tracking operations fired when updating index tables and reading unique indexes. This memory is used to store the key and column information for these operations. It is only very rarely that the value for this parameter needs to be altered from the default.
The default value for
TransactionBufferMemory
is 1MB.
Normal read and write operations use a similar buffer,
whose usage is even more short-lived. The compile-time
parameter ZATTRBUF_FILESIZE
(found in
ndb/src/kernel/blocks/Dbtc/Dbtc.hpp
)
set to 4000 脳 128 bytes (500KB). A similar buffer
for key information, ZDATABUF_FILESIZE
(also in Dbtc.hpp
) contains 4000
脳 16 = 62.5KB of buffer space.
Dbtc
is the module that handles
transaction coordination.
There are additional [NDBD]
parameters in
the Dblqh
module (in
ndb/src/kernel/blocks/Dblqh/Dblqh.hpp
)
that affect reads and updates. These include
ZATTRINBUF_FILESIZE
, set by default to
10000 脳 128 bytes (1250KB) and
ZDATABUF_FILE_SIZE
, set by default to
10000*16 bytes (roughly 156KB) of buffer space. To date,
there have been neither any reports from users nor any
results from our own extensive tests suggesting that either
of these compile-time limits should be increased.
This parameter is used to control the number of parallel
scans that can be performed in the cluster. Each
transaction coordinator can handle the number of parallel
scans defined for this parameter. Each scan query is
performed by scanning all partitions in parallel. Each
partition scan uses a scan record in the node where the
partition is located, the number of records being the
value of this parameter times the number of nodes. The
cluster should be able to sustain
MaxNoOfConcurrentScans
scans
concurrently from all nodes in the cluster.
Scans are actually performed in two cases. The first of these cases occurs when no hash or ordered indexes exists to handle the query, in which case the query is executed by performing a full table scan. The second case is encountered when there is no hash index to support the query but there is an ordered index. Using the ordered index means executing a parallel range scan. The order is kept on the local partitions only, so it is necessary to perform the index scan on all partitions.
The default value of
MaxNoOfConcurrentScans
is 256. The
maximum value is 500.
Specifies the number of local scan records if many scans
are not fully parallelized. If the number of local scan
records is not provided, it is calculated as the product
of MaxNoOfConcurrentScans
and the
number of data nodes in the system. The minimum value is
32.
This parameter is used to calculate the number of lock records which must be there to handle many concurrent scan operations.
The default value is 64; this value has a strong
connection to the ScanBatchSize
defined
in the SQL nodes.
This is an internal buffer used for passing messages within individual nodes and between nodes. Although it is highly unlikely that this would need to be changed, it is configurable. By default, it is set to 1MB.
These [NDBD]
parameters control log and
checkpoint behavior.
This parameter sets the number of REDO log files for the node, and thus the amount of space allocated to REDO logging. Because the REDO log files are organized in a ring, it is extremely important that the first and last log files in the set (sometimes referred to as the 鈥head鈥 and 鈥tail鈥 log files, respectively) do not meet. When these approach one another too closely, the node begins aborting all transactions encompassing updates due to a lack of room for new log records.
A REDO
log record is not removed until
three local checkpoints have been completed since that log
record was inserted. Checkpointing frequency is determined
by its own set of configuration parameters discussed
elsewhere in this chapter.
How these parameters interact and proposals for how to configure them are discussed in Section聽15.4.6, 鈥淐onfiguring Parameters for Local Checkpoints鈥.
The default parameter value is 8, which means 8 sets of 4
16MB files for a total of 512MB. In other words, REDO log
space must be allocated in blocks of 64MB. In scenarios
requiring a great many updates, the value for
NoOfFragmentLogFiles
may need to be set
as high as 300 or even higher to provide sufficient space
for REDO logs.
If the checkpointing is slow and there are so many writes
to the database that the log files are full and the log
tail cannot be cut without jeopardizing recovery, all
updating transactions are aborted with internal error code
410 (Out of log file space
temporarily
). This condition prevails until a
checkpoint has completed and the log tail can be moved
forward.
Important: This parameter
cannot be changed 鈥on the fly鈥; you must
restart the node using --initial
. If you
wish to change this value for a running cluster, you can
do so via a rolling node restart.
This parameter sets a ceiling on how many internal threads to allocate for open files. Any situation requiring a change in this parameter should be reported as a bug.
The default value is 40.
This parameter sets the initial number of internal threads to allocate for open files.
The default value is 27.
This parameter sets the maximum number of trace files that are kept before overwriting old ones. Trace files are generated when, for whatever reason, the node crashes.
The default is 25 trace files.
The next set of [NDBD]
parameters defines
pool sizes for metadata objects, used to define the maximum
number of attributes, tables, indexes, and trigger objects
used by indexes, events, and replication between clusters.
Note that these act merely as 鈥suggestions鈥 to
the cluster, and any that are not specified revert to the
default values shown.
Defines the number of attributes that can be defined in the cluster.
The default value is 1000, with the minimum possible value being 32. The maximum is 4294967039. Each attribute consumes around 200 bytes of storage per node due to the fact that all metadata is fully replicated on the servers.
When setting MaxNoOfAttributes
, it is
important to prepare in advance for any ALTER
TABLE
statements that you might want to perform
in the future. This is due to the fact, during the
execution of ALTER TABLE
on a Cluster
table, 3 times the number of attributes as in the original
table are used. For example, if a table requires 100
attributes, and you want to be able to alter it later, you
need to set the value of
MaxNoOfAttributes
to 300. Assuming that
you can create all desired tables without any problems, a
good rule of thumb is to add two times the number of
attributes in the largest table to
MaxNoOfAttributes
to be sure. You
should also verify that this number is sufficient by
trying an actual ALTER TABLE
after
configuring the parameter. If this is not successful,
increase MaxNoOfAttributes
by another
multiple of the original value and test it again.
A table object is allocated for each table, unique hash index, and ordered index. This parameter sets the maximum number of table objects for the cluster as a whole.
For each attribute that has a BLOB
data
type an extra table is used to store most of the
BLOB
data. These tables also must be
taken into account when defining the total number of
tables.
The default value of this parameter is 128. The minimum is 8 and the maximum is 1600. Each table object consumes approximately 20KB per node.
For each ordered index in the cluster, an object is allocated describing what is being indexed and its storage segments. By default, each index so defined also defines an ordered index. Each unique index and primary key has both an ordered index and a hash index.
The default value of this parameter is 128. Each object consumes approximately 10KB of data per node.
For each unique index that is not a primary key, a special
table is allocated that maps the unique key to the primary
key of the indexed table. By default, an ordered index is
also defined for each unique index. To prevent this, you
must specify the USING HASH
option when
defining the unique index.
The default value is 64. Each index consumes approximately 15KB per node.
Internal update, insert, and delete triggers are allocated for each unique hash index. (This means that three triggers are created for each unique hash index.) However, an ordered index requires only a single trigger object. Backups also use three trigger objects for each normal table in the cluster.
Replication between clusters also makes use of internal triggers.
This parameter sets the maximum number of trigger objects in the cluster.
The default value is 768.
This parameter is deprecated in MySQL 5.1;
you should use MaxNoOfOrderedIndexes
and MaxNoOfUniqueHashIndexes
instead.
This parameter is used only by unique hash indexes. There needs to be one record in this pool for each unique hash index defined in the cluster.
The default value of this parameter is 128.
The behavior of data nodes is also affected by a set of
[NDBD]
parameters taking on boolean values.
These parameters can each be specified as
TRUE
by setting them equal to
1
or Y
, and as
FALSE
by setting them equal to
0
or N
.
For a number of operating systems, including Solaris and Linux, it is possible to lock a process into memory and so avoid any swapping to disk. This can be used to help guarantee the cluster's real-time characteristics.
Beginning with MySQL 5.1.15, this parameter takes one of
the integer values 0
,
1
, or 2
, which act
as follows:
0
: Disables locking. This is the
default value.
1
: Performs the lock after
allocating memory for the process.
2
: Performs the lock before memory
for the process is allocated.
Previously, this parameter was a Boolean.
0
or false
was the
default setting, and disabled locking.
1
or true
enabled
locking of the process after its memory was allocated.
Important: Beginning with
MySQL 5.1.15, it is no longer possible to use
true
or false
for
the value of this parameter; when upgrading from a
previous version, you must change the value to
0
, 1
, or
2
.
This parameter specifies whether an ndbd process should exit or perform an automatic restart when an error condition is encountered.
This feature is enabled by default.
It is possible to specify MySQL Cluster tables as diskless, meaning that tables are not checkpointed to disk and that no logging occurs. Such tables exist only in main memory. A consequence of using diskless tables is that neither the tables nor the records in those tables survive a crash. However, when operating in diskless mode, it is possible to run ndbd on a diskless computer.
Important: This feature causes the entire cluster to operate in diskless mode.
When this feature is enabled, Cluster online backup is disabled. In addition, a partial start of the cluster is not possible.
Diskless
is disabled by default.
This feature is accessible only when building the debug version where it is possible to insert errors in the execution of individual blocks of code as part of testing.
This feature is disabled by default.
Controlling Timeouts, Intervals, and Disk Paging
There are a number of [NDBD]
parameters
specifying timeouts and intervals between various actions in
Cluster data nodes. Most of the timeout values are specified
in milliseconds. Any exceptions to this are mentioned where
applicable.
To prevent the main thread from getting stuck in an endless loop at some point, a 鈥watchdog鈥 thread checks the main thread. This parameter specifies the number of milliseconds between checks. If the process remains in the same state after three checks, the watchdog thread terminates it.
This parameter can easily be changed for purposes of experimentation or to adapt to local conditions. It can be specified on a per-node basis although there seems to be little reason for doing so.
The default timeout is 4000 milliseconds (4 seconds).
This parameter specifies how long the Cluster waits for all data nodes to come up before the cluster initialization routine is invoked. This timeout is used to avoid a partial Cluster startup whenever possible.
The default value is 30000 milliseconds (30 seconds). 0 disables the timeout, in which case the cluster may start only if all nodes are available.
If the cluster is ready to start after waiting for
StartPartialTimeout
milliseconds but is
still possibly in a partitioned state, the cluster waits
until this timeout has also passed.
The default timeout is 60000 milliseconds (60 seconds).
If a data node has not completed its startup sequence within the time specified by this parameter, the node startup fails. Setting this parameter to 0 (the default value) means that no data node timeout is applied.
For nonzero values, this parameter is measured in milliseconds. For data nodes containing extremely large amounts of data, this parameter should be increased. For example, in the case of a data node containing several gigabytes of data, a period as long as 10鈥15 minutes (that is, 600000 to 1000000 milliseconds) might be required to perform a node restart.
One of the primary methods of discovering failed nodes is by the use of heartbeats. This parameter states how often heartbeat signals are sent and how often to expect to receive them. After missing three heartbeat intervals in a row, the node is declared dead. Thus, the maximum time for discovering a failure through the heartbeat mechanism is four times the heartbeat interval.
The default heartbeat interval is 1500 milliseconds (1.5 seconds). This parameter must not be changed drastically and should not vary widely between nodes. If one node uses 5000 milliseconds and the node watching it uses 1000 milliseconds, obviously the node will be declared dead very quickly. This parameter can be changed during an online software upgrade, but only in small increments.
Each data node sends heartbeat signals to each MySQL
server (SQL node) to ensure that it remains in contact. If
a MySQL server fails to send a heartbeat in time it is
declared 鈥dead,鈥 in which case all ongoing
transactions are completed and all resources released. The
SQL node cannot reconnect until all activities initiated
by the previous MySQL instance have been completed. The
three-heartbeat criteria for this determination are the
same as described for
HeartbeatIntervalDbDb
.
The default interval is 1500 milliseconds (1.5 seconds). This interval can vary between individual data nodes because each data node watches the MySQL servers connected to it, independently of all other data nodes.
This parameter is an exception in that it does not specify a time to wait before starting a new local checkpoint; rather, it is used to ensure that local checkpoints are not performed in a cluster where relatively few updates are taking place. In most clusters with high update rates, it is likely that a new local checkpoint is started immediately after the previous one has been completed.
The size of all write operations executed since the start of the previous local checkpoints is added. This parameter is also exceptional in that it is specified as the base-2 logarithm of the number of 4-byte words, so that the default value 20 means 4MB (4 脳 220) of write operations, 21 would mean 8MB, and so on up to a maximum value of 31, which equates to 8GB of write operations.
All the write operations in the cluster are added
together. Setting
TimeBetweenLocalCheckpoints
to 6 or
less means that local checkpoints will be executed
continuously without pause, independent of the cluster's
workload.
When a transaction is committed, it is committed in main memory in all nodes on which the data is mirrored. However, transaction log records are not flushed to disk as part of the commit. The reasoning behind this behavior is that having the transaction safely committed on at least two autonomous host machines should meet reasonable standards for durability.
It is also important to ensure that even the worst of cases 鈥 a complete crash of the cluster 鈥 is handled properly. To guarantee that this happens, all transactions taking place within a given interval are put into a global checkpoint, which can be thought of as a set of committed transactions that has been flushed to disk. In other words, as part of the commit process, a transaction is placed in a global checkpoint group. Later, this group's log records are flushed to disk, and then the entire group of transactions is safely committed to disk on all computers in the cluster.
This parameter defines the interval between global checkpoints. The default is 2000 milliseconds.
TimeBetweenInactiveTransactionAbortCheck
Timeout handling is performed by checking a timer on each transaction once for every interval specified by this parameter. Thus, if this parameter is set to 1000 milliseconds, every transaction will be checked for timing out once per second.
The default value is 1000 milliseconds (1 second).
This parameter states the maximum time that is permitted to lapse between operations in the same transaction before the transaction is aborted.
The default for this parameter is zero (no timeout). For a real-time database that needs to ensure that no transaction keeps locks for too long, this parameter should be set to a much smaller value. The unit is milliseconds.
TransactionDeadlockDetectionTimeout
When a node executes a query involving a transaction, the node waits for the other nodes in the cluster to respond before continuing. A failure to respond can occur for any of the following reasons:
The node is 鈥dead鈥
The operation has entered a lock queue
The node requested to perform the action could be heavily overloaded.
This timeout parameter states how long the transaction coordinator waits for query execution by another node before aborting the transaction, and is important for both node failure handling and deadlock detection. Setting it too high can cause a undesirable behavior in situations involving deadlocks and node failure.
The default timeout value is 1200 milliseconds (1.2 seconds).
This is the maximum number of bytes to store before flushing data to a local checkpoint file.
The default value is 4M (4 megabytes).
This parameter was added in MySQL 5.1.12.
The amount of data,in bytes per second, that is sent to disk during a local checkpoint.
The default value is 10M (10 megabytes per second).
This parameter was added in MySQL 5.1.12.
The amount of data,in bytes per second, that is sent to disk during a local checkpoint as part of a restart operation.
The default value is 100M (100 megabytes per second).
This parameter was added in MySQL 5.1.12.
NoOfDiskPagesToDiskAfterRestartTUP
When executing a local checkpoint, the algorithm flushes
all data pages to disk. Merely doing so as quickly as
possible without any moderation is likely to impose
excessive loads on processors, networks, and disks. To
control the write speed, this parameter specifies how many
pages per 100 milliseconds are to be written. In this
context, a 鈥page鈥 is defined as 8KB. This
parameter is specified in units of 80KB per second, so ,
setting
NoOfDiskPagesToDiskAfterRestartTUP
to a
value of 20
entails writing 1.6MB in
data pages to disk each second during a local checkpoint.
That is, this parameter handles the limitation of writes
from data memory. (See the entry for
IndexMemory
for information about index
pages.)
In short, this parameter specifies how quickly to execute
local checkpoints. It operates in conjunction with
NoOfFragmentLogFiles
,
DataMemory
, and
IndexMemory
.
For more information about the interaction between these parameters and possible strategies for choosing appropriate values for them, see Section聽15.4.6, 鈥淐onfiguring Parameters for Local Checkpoints鈥.
The default value is 40 (3.2MB of data pages per second).
Note: This parameter is deprecated as of MySQL 5.1.6. For MySQL 5.1.12 and later versions, use DiskCheckpointSpeed and DiskSyncSize.
NoOfDiskPagesToDiskAfterRestartACC
This parameter uses the same units as
NoOfDiskPagesToDiskAfterRestartTUP
and
acts in a similar fashion, but limits the speed of writing
index pages from index memory.
The default value of this parameter is 20 (1.6MB of index memory pages per second).
Note: This parameter is deprecated as of MySQL 5.1.6. For MySQL 5.1.12 and later versions, use DiskCheckpointSpeed and DiskSyncSize.
NoOfDiskPagesToDiskDuringRestartTUP
This parameter is used in a fashion similar to
NoOfDiskPagesToDiskAfterRestartTUP
and
NoOfDiskPagesToDiskAfterRestartACC
,
only it does so with regard to local checkpoints executed
in the node when a node is restarting. A local checkpoint
is always performed as part of all node restarts. During a
node restart it is possible to write to disk at a higher
speed than at other times, because fewer activities are
being performed in the node.
This parameter covers pages written from data memory.
The default value is 40 (3.2MB per second).
Note: This parameter is deprecated as of MySQL 5.1.6. For MySQL 5.1.12 and later versions, use DiskCheckpointSpeedInRestart and DiskSyncSize.
NoOfDiskPagesToDiskDuringRestartACC
Controls the number of index memory pages that can be written to disk during the local checkpoint phase of a node restart.
As with
NoOfDiskPagesToDiskAfterRestartTUP
and
NoOfDiskPagesToDiskAfterRestartACC
,
values for this parameter are expressed in terms of 8KB
pages written per 100 milliseconds (80KB/second).
The default value is 20 (1.6MB per second).
Note: This parameter is deprecated as of MySQL 5.1.6. For MySQL 5.1.12 and later versions, use DiskCheckpointSpeedInRestart and DiskSyncSize.
This parameter specifies how long data nodes wait for a response from the arbitrator to an arbitration message. If this is exceeded, the network is assumed to have split.
The default value is 1000 milliseconds (1 second).
All update activities need to be logged. The REDO log makes it possible to replay these updates whenever the system is restarted. The NDB recovery algorithm uses a 鈥fuzzy鈥 checkpoint of the data, and then applies the REDO log to play back all changes up to the restoration point.
RedoBuffer
sets the size of the buffer
inwhich the REDO log is written, and is 8MB by default. This
buffer is used as a front end to the file system when writing
REDO log records to disk. If this buffer is too small, the
NDB
storage engine issues error code 1221
(REDO log buffers overloaded
).
The minimum value is 1MB. If the node is running in diskless
mode, these parameters can be set to their minimum values
without penalty due to the fact that disk writes are
鈥faked鈥 by the NDB
storage
engine's filesystem abstraction layer.
Important: It is not safe to decrease the value of this parameter during a rolling restart.
Note: The
UndoIndexBuffer
and
UndoDataBuffer
parameters which appeared in
earlier versions of MySQL Cluster are not necessary in MySQL
5.1, since undo logs are no longer kept on disk for in-memory
tables.
In managing the cluster, it is very important to be able to
control the number of log messages sent for various event
types to stdout
. For each event category,
there are 16 possible event levels (numbered 0 through 15).
Setting event reporting for a given event category to level 15
means all event reports in that category are sent to
stdout
; setting it to 0 means that there
will be no event reports made in that category.
By default, only the startup message is sent to
stdout
, with the remaining event reporting
level defaults being set to 0. The reason for this is that
these messages are also sent to the management server's
cluster log.
An analogous set of levels can be set for the management client to determine which event levels to record in the cluster log.
The reporting level for events generated during startup of the process.
The default level is 1.
The reporting level for events generated as part of graceful shutdown of a node.
The default level is 0.
The reporting level for statistical events such as number of primary key reads, number of updates, number of inserts, information relating to buffer usage, and so on.
The default level is 0.
The reporting level for events generated by local and global checkpoints.
The default level is 0.
The reporting level for events generated during node restart.
The default level is 0.
The reporting level for events generated by connections between cluster nodes.
The default level is 0.
The reporting level for events generated by errors and warnings by the cluster as a whole. These errors do not cause any node failure but are still considered worth reporting.
The default level is 0.
The reporting level for events generated by congestion. These errors do not cause node failure but are still considered worth reporting.
The default level is 0.
The reporting level for events generated for information about the general state of the cluster.
The default level is 0.
This parameter controls how often data node memory usage reports are recorded in the cluster log; it is an integer value representing the number of seconds between reports.
Each data node's data memory and index memory usage is
logged as both a percentage and a number of 32 KB pages of
the DataMemory
and
IndexMemory
, respectively, set in the
config.ini
file. For example, if
DataMemory
is equal to 100 MB, and a
given data node is using 50 MB for data memory storage,
the corresponding line in the cluster log might look like
this:
2006-12-24 01:18:16 [MgmSrvr] INFO -- Node 2: Data usage is 50%(1280 32K pages of total 2560)
MemReportFrequency
is not a required
parameter. If used, it can be set for all cluster data
nodes in the [NDBD DEFAULT]
section of
config.ini
, and can also be set or
overridden for individual data nodes in the corresponding
[NDBD]
sections of the configuration
file. The minimum value 鈥 which is also the default
value 鈥 is 0, in which case memory reports are
logged only when memory usage reaches certain percentages
(80%, 90%, and 100%), as mentioned in the discussion of
statistics events in
Section聽15.7.3.2, 鈥淟og Events鈥.
This parameter was added in MySQL Cluster 5.1.16.
The [NDBD]
parameters discussed in this
section define memory buffers set aside for execution of
online backups.
In creating a backup, there are two buffers used for
sending data to the disk. The backup data buffer is used
to fill in data recorded by scanning a node's tables. Once
this buffer has been filled to the level specified as
BackupWriteSize
(see below), the pages
are sent to disk. While flushing data to disk, the backup
process can continue filling this buffer until it runs out
of space. When this happens, the backup process pauses the
scan and waits until some disk writes have completed freed
up memory so that scanning may continue.
The default value is 2MB.
The backup log buffer fulfills a role similar to that played by the backup data buffer, except that it is used for generating a log of all table writes made during execution of the backup. The same principles apply for writing these pages as with the backup data buffer, except that when there is no more space in the backup log buffer, the backup fails. For that reason, the size of the backup log buffer must be large enough to handle the load caused by write activities while the backup is being made. See Section聽15.8.4, 鈥淐onfiguration for Cluster Backup鈥.
The default value for this parameter should be sufficient for most applications. In fact, it is more likely for a backup failure to be caused by insufficient disk write speed than it is for the backup log buffer to become full. If the disk subsystem is not configured for the write load caused by applications, the cluster is unlikely to be able to perform the desired operations.
It is preferable to configure cluster nodes in such a manner that the processor becomes the bottleneck rather than the disks or the network connections.
The default value is 2MB.
This parameter is simply the sum of
BackupDataBufferSize
and
BackupLogBufferSize
.
The default value is 2MB + 2MB = 4MB.
Important: If
BackupDataBufferSize
and
BackupLogBufferSize
taken together
exceed 4MB, then this parameter must be set explicitly in
the config.ini
file to their sum.
This parameter specifies the default size of messages written to disk by the backup log and backup data buffers.
The default value is 32KB.
This parameter specifies the maximum size of messages written to disk by the backup log and backup data buffers.
The default value is 256KB.
Important: When specifying these parameters, the following relationships must hold true. Otherwise, the data node will be unable to start.
BackupDataBufferSize >= BackupWriteSize +
188KB
BackupLogBufferSize >= BackupWriteSize +
16KB
BackupMaxWriteSize >=
BackupWriteSize
The [MYSQLD]
and [API]
sections in the config.ini
file define
the behavior of the MySQL servers (SQL nodes) and other
applications (API nodes) used to access cluster data. None of
the parameters shown is required. If no computer or host name
is provided, any host can use this SQL or API node.
Generally speaking, a [MYSQLD]
section is
used to indicate a MySQL server providing an SQL interface to
the cluster, and an [API]
section is used
for applications other than mysqld
processes accessing cluster data, but the two designations are
actually synonomous; you can, for instance, list parameters
for a MySQL server acting as an SQL node in an
[API]
section.
The Id
value is used to identify the
node in all cluster internal messages. It must be an
integer in the range 1 to 63 inclusive, and must be unique
among all node IDs within the cluster.
This refers to the Id
set for one of
the computers (hosts) defined in a
[COMPUTER]
section of the configuration
file.
Specifying this parameter defines the hostname of the
computer on which the SQL node (API node) is to reside. To
specify a hostname other than
localhost
, either this parameter or
ExecuteOnComputer
is required.
This parameter defines which nodes can act as arbitrators.
Both MGM nodes and SQL nodes can be arbitrators. A value
of 0 means that the given node is never used as an
arbitrator, a value of 1 gives the node high priority as
an arbitrator, and a value of 2 gives it low priority. A
normal configuration uses the management server as
arbitrator, setting its ArbitrationRank
to 1 (the default) and those for all SQL nodes to 0.
Setting this parameter to any other value than 0 (the default) means that responses by the arbitrator to arbitration requests will be delayed by the stated number of milliseconds. It is usually not necessary to change this value.
For queries that are translated into full table scans or
range scans on indexes, it is important for best
performance to fetch records in properly sized batches. It
is possible to set the proper size both in terms of number
of records (BatchSize
) and in terms of
bytes (BatchByteSize
). The actual batch
size is limited by both parameters.
The speed at which queries are performed can vary by more than 40% depending upon how this parameter is set. In future releases, MySQL Server will make educated guesses on how to set parameters relating to batch size, based on the query type.
This parameter is measured in bytes and by default is equal to 32KB.
This parameter is measured in number of records and is by default set to 64. The maximum size is 992.
The batch size is the size of each batch sent from each data node. Most scans are performed in parallel to protect the MySQL Server from receiving too much data from many nodes in parallel; this parameter sets a limit to the total batch size over all nodes.
The default value of this parameter is set to 256KB. Its maximum size is 16MB.
You can obtain some information from a MySQL server running as
a Cluster SQL node using SHOW STATUS
in the
mysql
client, as shown here:
mysql> SHOW STATUS LIKE 'ndb%';
+-----------------------------+---------------+
| Variable_name | Value |
+-----------------------------+---------------+
| Ndb_cluster_node_id | 5 |
| Ndb_config_from_host | 192.168.0.112 |
| Ndb_config_from_port | 1186 |
| Ndb_number_of_storage_nodes | 4 |
+-----------------------------+---------------+
4 rows in set (0.02 sec)
For information about these Cluster system status variables, see Section聽5.2.5, 鈥淪tatus Variables鈥.
TCP/IP is the default transport mechanism for establishing
connections in MySQL Cluster. It is normally not necessary to
define connections because Cluster automatically set ups a
connection between each of the data nodes, between each data
node and all MySQL server nodes, and between each data node
and the management server. (For one exception to this rule,
see Section聽15.4.4.8, 鈥淭CP/IP Connections Using Direct Connections鈥.)
[TCP]
sections in the
config.ini
file explicitly define TCP/IP
connections between nodes in the cluster.
It is necessary to define a connection only to override the
default connection parameters. In that case, it is necessary
to define at least NodeId1
,
NodeId2
, and the parameters to change.
Any [TCP]
sections in the
config.ini
file should be listed last,
following any other sections in the file. This is not
required for a [TCP DEFAULT]
section.
This is a known issue with the way in which the
config.ini
file is read by the cluster
management server.
It is also possible to change the default values for these
parameters by setting them in the [TCP
DEFAULT]
section.
To identify a connection between two nodes it is necessary
to provide their node IDs in the [TCP]
section of the configuration file. These are the same
unique Id
values for each of these
nodes as described in
Section聽15.4.4.6, 鈥淒efining SQL and Other API Nodes鈥.
TCP transporters use a buffer to store all messages before performing the send call to the operating system. When this buffer reaches 64KB its contents are sent; these are also sent when a round of messages have been executed. To handle temporary overload situations it is also possible to define a bigger send buffer. The default size of the send buffer is 256KB.
To be able to retrace a distributed message datagram, it
is necessary to identify each message. When this parameter
is set to Y
, message IDs are
transported over the network. This feature is disabled by
default in production builds, and enabled in
-debug
builds.
This parameter is a boolean parameter (enabled by setting
it to Y
or 1
,
disabled by setting it to N
or
0
). It is disabled by default. When it
is enabled, checksums for all messages are calculated
before they placed in the send buffer. This feature
ensures that messages are not corrupted while waiting in
the send buffer, or by the transport mechanism.
This formerly specified the port number to be used for listening for connections from other nodes. This parameter should no longer be used.
Specifies the size of the buffer used when receiving data from the TCP/IP socket. There is seldom any need to change this parameter from its default value of 64KB, except possibly to save memory.
Setting up a cluster using direct connections between data
nodes requires specifying explicitly the crossover IP
addresses of the data nodes so connected in the
[TCP]
section of the cluster
config.ini
file.
In the following example, we envision a cluster with at least
four hosts, one each for a management server, an SQL node, and
two data nodes. The cluster as a whole resides on the
172.23.72.*
subnet of a LAN. In addition to
the usual network connections, the two data nodes are
connected directly using a standard crossover cable, and
communicate with one another directly using IP addresses in
the 1.1.0.*
address range as shown:
# Management Server [NDB_MGMD] Id=1 HostName=172.23.72.20 # SQL Node [MYSQLD] Id=2 HostName=172.23.72.21 # Data Nodes [NDBD] Id=3 HostName=172.23.72.22 [NDBD] Id=4 HostName=172.23.72.23 # TCP/IP Connections [TCP] NodeId1=3 NodeId2=4 HostName1=1.1.0.1 HostName2=1.1.0.2
The HostName
parameter, where N
N
is an integer,
is used only when specifying direct TCP/IP connections.
The use of direct connections between data nodes can improve the cluster's overall efficiency by allowing the data nodes to bypass an Ethernet device such as a switch, hub, or router, thus cutting down on the cluster's latency. It is important to note that to take the best advantage of direct connections in this fashion with more than two data nodes, you must have a direct connection between each data node and every other data node in the same node group.
MySQL Cluster attempts to use the shared memory transporter
and configure it automatically where possible.
[SHM]
sections in the
config.ini
file explicitly define
shared-memory connections between nodes in the cluster. When
explicitly defining shared memory as the connection method, it
is necessary to define at least NodeId1
,
NodeId2
and ShmKey
. All
other parameters have default values that should work well in
most cases.
Important: SHM functionality is considered experimental only. It is not officially supported in any MySQL release series up to and including 5.1. This means that you must determine for yourself or by using our free resources (forums, mailing lists) whether it can be made to work correctly in your specific case.
To identify a connection between two nodes it is necessary
to provide node identifiers for each of them, as
NodeId1
and NodeId2
.
When setting up shared memory segments, a node ID, expressed as an integer, is used to identify uniquely the shared memory segment to use for the communication. There is no default value.
Each SHM connection has a shared memory segment where
messages between nodes are placed by the sender and read
by the reader. The size of this segment is defined by
ShmSize
. The default value is 1MB.
To retrace the path of a distributed message, it is
necessary to provide each message with a unique
identifier. Setting this parameter to Y
causes these message IDs to be transported over the
network as well. This feature is disabled by default in
production builds, and enabled in
-debug
builds.
This parameter is a boolean
(Y
/N
) parameter
which is disabled by default. When it is enabled,
checksums for all messages are calculated before being
placed in the send buffer.
This feature prevents messages from being corrupted while waiting in the send buffer. It also serves as a check against data being corrupted during transport.
[SCI]
sections in the
config.ini
file explicitly define SCI
(Scalable Coherent Interface) connections between cluster
nodes. Using SCI transporters in MySQL Cluster is supported
only when the MySQL binaries are built using
--with-ndb-sci=
.
The /your/path/to/SCI
path
should point to a
directory that contains at a minimum lib
and include
directories containing SISCI
libraries and header files. (See
Section聽15.12, 鈥淯sing High-Speed Interconnects with MySQL Cluster鈥 for more
information about SCI.)
In addition, SCI requires specialized hardware.
It is strongly recommended to use SCI Transporters only for communication between ndbd processes. Note also that using SCI Transporters means that the ndbd processes never sleep. For this reason, SCI Transporters should be used only on machines having at least two CPUs dedicated for use by ndbd processes. There should be at least one CPU per ndbd process, with at least one CPU left in reserve to handle operating system activities.
To identify a connection between two nodes it is necessary
to provide node identifiers for each of them, as
NodeId1
and NodeId2
.
This identifies the SCI node ID on the first Cluster node
(identified by NodeId1
).
It is possible to set up SCI Transporters for failover between two SCI cards which then should use separate networks between the nodes. This identifies the node ID and the second SCI card to be used on the first node.
This identifies the SCI node ID on the second Cluster node
(identified by NodeId2
).
When using two SCI cards to provide failover, this parameter identifies the second SCI card to be used on the second node.
Each SCI transporter has a shared memory segment used for communication between the two nodes. Setting the size of this segment to the default value of 1MB should be sufficient for most applications. Using a smaller value can lead to problems when performing many parallel inserts; if the shared buffer is too small, this can also result in a crash of the ndbd process.
A small buffer in front of the SCI media stores messages before transmitting them over the SCI network. By default, this is set to 8KB. Our benchmarks show that performance is best at 64KB but 16KB reaches within a few percent of this, and there was little if any advantage to increasing it beyond 8KB.
To trace a distributed message it is necessary to identify
each message uniquely. When this parameter is set to
Y
, message IDs are transported over the
network. This feature is disabled by default in production
builds, and enabled in -debug
builds.
This parameter is a boolean value, and is disabled by
default. When Checksum
is enabled,
checksums are calculated for all messages before they are
placed in the send buffer. This feature prevents messages
from being corrupted while waiting in the send buffer. It
also serves as a check against data being corrupted during
transport.
The next three sections provide summary tables of MySQL Cluster
configuration parameters used in the
config.ini
file to govern the cluster's
functioning. Each table lists the parameters for one of the
Cluster node process types (ndbd,
ndb_mgmd, and mysqld), and
includes the parameter's type as well as its default, mimimum,
and maximum values as applicable.
It is also stated what type of restart is required (node restart
or system restart) 鈥 and whether the restart must be done
with --initial
鈥 to change the value of a
given configuration parameter. This information is provided in
each table's Restart Type
column, which contains one of the values shown in this list:
N
: Node Restart
IN
: Initial Node Restart
S
: System Restart
IS
: Initial System Restart
When performing a node restart or an initial node restart, all
of the cluster's data nodes must be restarted in turn (also
referred to as a rolling restart). It is
possible to update cluster configuration parameters marked
N
or IN
online 鈥
that is, without shutting down the cluster 鈥 in this
fashion. An initial node restart requires restarting each
ndbd process with the
--initial
option.
A system restart requires a complete shutdown and restart of the entire cluster. An initial system restart requires taking a backup of the cluster, wiping the cluster filesystem after shutdown, and then restoring from the backup following the restart.
In any cluster restart, all of the cluster's management servers must be restarted in order for them to read the updated configuration parameter values.
Important: Values for numeric cluster parameters can generally be increased without any problems, although it is advisable to do so progressively, making such adjustments in relatively small increments. However, decreasing the values of such parameters 鈥 particularly those relating to memory usage and disk space 鈥 is not to be undertaken lightly, and it is recommended that you do so only following careful planning and testing. In addition, it is the generally the case that parameters relating to memory and disk usage which can be raised using a simple node restart require an initial node restart to be lowered.
Because some of these parameters can be used for configuring more than one type of cluster node, they may appear in more than one of the tables.
(Note that 4294967039
鈥 which often
appears as a maximum value in these tables 鈥 is equal to
232 鈥
28 鈥 1
.)
The following table provides information about parameters used
in the [NDBD]
or
[NDB_DEFAULT]
sections of a
config.ini
file for configuring MySQL
Cluster data nodes. For detailed descriptions and other
additional information about each of these parameters, see
Section聽15.4.4.5, 鈥淒efining Data Nodes鈥.
Restart Type Column Values
N
: Node Restart
IN
: Initial Node Restart
S
: System Restart
IS
: Initial System Restart
See Section聽15.4.5, 鈥淥verview of Cluster Configuration Parameters鈥, for additional explanations of these abbreviations.
Parameter Name | Type/Units | Default Value | Minimum Value | Maximum Value | Restart Type |
ArbitrationTimeout | milliseconds | 3000 | 10 | 4294967039 | N |
BackupDataBufferSize | bytes | 2M | 0 | 4294967039 | N |
BackupDataDir | string |
| N/A | N/A | IN |
BackupLogBufferSize | bytes | 2M | 0 | 4294967039 | N |
BackupMemory | bytes | 4M | 0 | 4294967039 | N |
BackupWriteSize | bytes | 32K | 2K | 4294967039 | N |
BackupMaxWriteSize | bytes | 256K | 2K | 4294967039 | N |
BatchSizePerLocalScan | integer | 64 | 1 | 992 | N |
DataDir | string | /var/lib/mysql-cluster | N/A | N/A | IN |
DataMemory | bytes | 80M | 1M | 1024G (subject to available system RAM and size of
IndexMemory ) | N |
DiskCheckpointSpeed
(added in MySQL 5.1.12) | integer (number of bytes per second) | 10M | 1M | 4294967039 | N |
DiskCheckpointSpeedInRestart
(added in MySQL 5.1.12) | integer (number of bytes per second) | 100M | 1M | 4294967039 | N |
Diskless | true|false (1 |0 ) | 0 | 0 | 1 | IS |
DiskPageBufferMemory | bytes | 64M | 4M | 1024G | IS |
DiskSyncSize
(added in MySQL 5.1.12) | integer (number of bytes) | 4M | 32K | 4294967039 | N |
ExecuteOnComputer | integer | 聽 | 聽 | 聽 | 聽 |
FileSystemPath | string | value specified for DataDir | N/A | N/A | IN |
HeartbeatIntervalDbApi | milliseconds | 1500 | 100 | 4294967039 | N |
HeartbeatIntervalDbDb | milliseconds | 1500 | 10 | 4294967039 | N |
HostName | string | localhost | N/A | N/A | S |
Id | integer | None | 1 | 63 | N |
IndexMemory | bytes | 18M | 1M | 1024G (subject to available system RAM and size of
DataMemory ) | N |
InitialNoOfOpenFiles | integer | 27 | 20 | 4294967039 | N |
LockPagesInMainMemory | As of MySQL 5.1.15: integer;
previously: true|false
(1 |0 ) | 0 | 0 | 1 | N |
LogLevelCheckpoint | integer | 0 | 0 | 15 | IN |
LogLevelCongestion | integer | 0 | 0 | 15 | N |
LogLevelConnection | integer | 0 | 0 | 15 | N |
LogLevelError | integer | 0 | 0 | 15 | N |
LogLevelInfo | integer | 0 | 0 | 15 | N |
LogLevelNodeRestart | integer | 0 | 0 | 15 | N |
LogLevelShutdown | integer | 0 | 0 | 15 | N |
LogLevelStartup | integer | 1 | 0 | 15 | N |
LogLevelStatistic | integer | 0 | 0 | 15 | N |
LongMessageBuffer | bytes | 1M | 512K | 4294967039 | N |
MaxNoOfAttributes | integer | 1000 | 32 | 4294967039 | N |
MaxNoOfConcurrentIndexOperations | integer | 8K | 0 | 4294967039 | N |
MaxNoOfConcurrentOperations | integer | 32768 | 32 | 4294967039 | N |
MaxNoOfConcurrentScans | integer | 256 | 2 | 500 | N |
MaxNoOfConcurrentTransactions | integer | 4096 | 32 | 4294967039 | N |
MaxNoOfFiredTriggers | integer | 4000 | 0 | 4294967039 | N |
MaxNoOfIndexes
(DEPRECATED 鈥 use
MaxNoOfOrderedIndexes or
MaxNoOfUniqueHashIndexes instead) | integer | 128 | 0 | 4294967039 | N |
MaxNoOfLocalOperations | integer | UNDEFINED | 32 | 4294967039 | N |
MaxNoOfLocalScans | integer | UNDEFINED (see
description) | 32 | 4294967039 | N |
MaxNoOfOpenFiles | integer | 40 | 20 | 4294967039 | N |
MaxNoOfOrderedIndexes | integer | 128 | 0 | 4294967039 | N |
MaxNoOfSavedMessages | integer | 25 | 0 | 4294967039 | N |
MaxNoOfTables | integer | 128 | 8 | 4294967039 | N |
MaxNoOfTriggers | integer | 768 | 0 | 4294967039 | N |
MaxNoOfUniqueHashIndexes | integer | 64 | 0 | 4294967039 | N |
MemReportFrequency
(added in MySQL 5.1.16) | integer | 0 | 0 | 4294967039 | N |
NoOfDiskPagesToDiskAfterRestartACC
(DEPRECATED as of MySQL 5.1.6) | integer (number of 8KB pages per 100 milliseconds) | 20 (= 20 * 80KB = 1.6MB/second) | 1 | 4294967039 | N |
NoOfDiskPagesToDiskAfterRestartTUP
(DEPRECATED as of MySQL 5.1.6) | integer (number of 8KB pages per 100 milliseconds) | 40 (= 40 * 80KB = 3.2MB/second) | 1 | 4294967039 | N |
NoOfDiskPagesToDiskDuringRestartACC
(DEPRECATED as of MySQL 5.1.6) | integer (number of 8KB pages per 100 milliseconds) | 20 (= 20 * 80KB = 1.6MB/second) | 1 | 4294967039 | N |
NoOfDiskPagesToDiskDuringRestartTUP
(DEPRECATED as of MySQL 5.1.6) | integer (number of 8KB pages per 100 milliseconds) | 40 (= 40 * 80KB = 3.2MB/second) | 1 | 4294967039 | N |
NoOfFragmentLogFiles | integer | 16 | 3 | 4294967039 | IN |
NoOfReplicas | integer | None | 1 | 4 | IS |
RedoBuffer | bytes | 8M | 1M | 4294967039 | N |
RestartOnErrorInsert
(DEBUG BUILDS ONLY) | true|false (1 |0 ) | 0 | 0 | 1 | N |
ServerPort
(OBSOLETE) | integer | 1186 | 0 | 4294967039 | N |
SharedGlobalmemory | bytes | 20M | 0 | 65536G | N |
StartFailureTimeout | milliseconds | 0 | 0 | 4294967039 | N |
StartPartialTimeout | milliseconds | 30000 | 0 | 4294967039 | N |
StartPartitionedTimeout | milliseconds | 60000 | 0 | 4294967039 | N |
StopOnError | true|false (1 |0 ) | 1 | 0 | 1 | N |
StringMemory | integer or percentage (see description for details) | 0 | 0 | 4294967039 | S |
TimeBetweenGlobalCheckpoints | milliseconds | 2000 | 10 | 32000 | N |
TimeBetweenInactiveTransactionAbortCheck | milliseconds | 1000 | 1000 | 4294967039 | N |
TimeBetweenLocalCheckpoints | integer (number of 4-byte words as a base-2 logarithm) | 20 (= 4 * 220 = 4MB write
operations) | 0 | 31 | N |
TimeBetweenWatchDogCheck | milliseconds | 6000 | 70 | 4294967039 | N |
TransactionBufferMemory | bytes | 1M | 1K | 4294967039 | N |
TransactionDeadlockDetectionTimeout | milliseconds | 1200 | 50 | 4294967039 | N |
TransactionInactiveTimeout | milliseconds | 0 | 0 | 4294967039 | N |
UndoDataBuffer
(OBSOLETE) | bytes | 2M | 1M | 4294967039 | N |
UndoDataBuffer
(OBSOLETE) | bytes | 2M | 1M | 4294967039 | N |
The following table provides information about parameters used
in the [NDB_MGMD]
or
[MGM]
sections of a
config.ini
file for configuring MySQL
Cluster management nodes. For detailed descriptions and other
additional information about each of these parameters, see
Section聽15.4.4.4, 鈥淒efining the Management Server鈥.
Restart Type Column Values
N
: Node Restart
IN
: Initial Node Restart
S
: System Restart
IS
: Initial System Restart
See Section聽15.4.5, 鈥淥verview of Cluster Configuration Parameters鈥, for additional explanations of these abbreviations.
Parameter Name | Type/Units | Default Value | Minimum Value | Maximum Value | Restart Type |
ArbitrationDelay | milliseconds | 0 | 0 | 4294967039 | N |
ArbitrationRank | integer | 1 | 0 | 2 | N |
DataDir | string | ./ (ndb_mgmd directory) | N/A | N/A | IN |
ExecuteOnComputer | integer | 聽 | 聽 | 聽 | 聽 |
HostName | string | localhost | N/A | N/A | IN |
Id | integer | None | 1 | 63 | IN |
LogDestination | CONSOLE , SYSLOG , or
FILE | FILE (see
Section聽15.4.4.4, 鈥淒efining the Management Server鈥) | N/A | N/A | N |
The following table provides information about parameters used
in the [SQL]
and [API]
sections of a config.ini
file for
configuring MySQL Cluster SQL nodes and API nodes. For
detailed descriptions and other additional information about
each of these parameters, see
Section聽15.4.4.6, 鈥淒efining SQL and Other API Nodes鈥.
Restart Type Column Values
N
: Node Restart
IN
: Initial Node Restart
S
: System Restart
IS
: Initial System Restart
See Section聽15.4.5, 鈥淥verview of Cluster Configuration Parameters鈥, for additional explanations of these abbreviations.
Parameter Name | Type/Units | Default Value | Minimum Value | Maximum Value | Restart Type |
ArbitrationDelay | milliseconds | 0 | 0 | 4294967039 | N |
ArbitrationRank | integer | 1 | 0 | 2 | N |
BatchByteSize | bytes | 32K | 1K | 1M | N |
BatchSize | integer | 64 | 1 | 992 | N |
ExecuteOnComputer | integer | 聽 | 聽 | 聽 | 聽 |
HostName | string | localhost | N/A | N/A | IN |
Id | integer | None | 1 | 63 | IN |
MaxScanBatchSize | bytes | 256K | 32K | 16M | N |
The parameters discussed in
Logging
and Checkpointing and in
Data
Memory, Index Memory, and String Memory that are used to
configure local checkpoints for a MySQL Cluster do not exist in
isolation, but rather are very much interdepedent on each other.
In this section, we illustrate how these parameters 鈥
including DataMemory
,
IndexMemory
,
NoOfDiskPagesToDiskAfterRestartTUP
,
NoOfDiskPagesToDiskAfterRestartACC
, and
NoOfFragmentLogFiles
鈥 relate to one
another in a working Cluster.
Important: The parameters
NoOfDiskPagesToDiskAfterRestartTUP
and
NoOfDiskPagesToDiskAfterRestartACC
were
deprecated in MySQL 5.1.6. From MySQL 5.1.6 through 5.1.11, disk
writes during LCPs took place at the maximum speed possible.
Beginning with MySQL 5.1.12, the speed and throughput for LCPs
are controlled using the parameters
DiskSyncSize
,
DiskCheckpointSpeed
, and
DiskCheckpointSpeedInRestart
. See
Section聽15.4.4.5, 鈥淒efining Data Nodes鈥.
In this example, we assume that our application performs the following numbers of types of operations per hour:
50000 selects
15000 inserts
15000 updates
15000 deletes
We also make the following assumptions about the data used in the application:
We are working with a single table having 40 columns.
Each column can hold up to 32 bytes of data.
A typical UPDATE
run by the application
affects the values of 5 columns.
No NULL
values are inserted by the
application.
A good starting point is to determine the amount of time that should elapse between local checkpoints (LCPs). It worth noting that, in the event of a system restart, it takes 40-60 percent of this interval to execute the REDO log 鈥 for example, if the time between LCPs is 5 minutes (300 seconds), then it should take 2 to 3 minutes (120 to 180 seconds) for the REDO log to be read.
The maximum amount of data per node can be assumed to be the
size of the DataMemory
parameter. In this
example, we assume that this is 2 GB. The
NoOfDiskPagesToDiskAfterRestartTUP
parameter
represents the amount of data to be checkpointed per unit time
鈥 however, this parameter is actually expressed as the
number of 8K memory pages to be checkpointed per 100
milliseconds. 2 GB per 300 seconds is approximately 6.8 MB per
second, or 700 KB per 100 milliseconds, which works out to
roughly 85 pages per 100 milliseconds.
Similarly, we can calculate
NoOfDiskPagesToDiskAfterRestartACC
in terms
of the time for local checkpoints and the amount of memory
required for indexes 鈥 that is, the
IndexMemory
. Assuming that we allow 512 MB
for indexes, this works out to approximately 20 8-KB pages per
100 milliseconds for this parameter.
Next, we need to determine the number of REDO log files required
鈥 that is, fragment log files 鈥 the corresponding
parameter being NoOfFragmentLogFiles
. We need
to make sure that there are sufficient REDO log files for
keeping records for at least 3 local checkpoints. In a
production setting, there are always uncertainties 鈥 for
instance, we cannot be sure that disks always operate at top
speed or with maximum throughput. For this reason, it is best to
err on the side of caution, so we double our requirement and
calculate a number of fragment log files which should be enough
to keep records covering 6 local checkpoints.
It is also important to remember that the disk also handles
writes to the REDO log, so if you find that the amount of data
being written to disk as detemined by the values of
NoOfDiskPagesToDiskAfterRestartACC
and
NoOfDiskPagesToDiskAfterRestartTUP
is
approaching the amount of disk bandwidth available, you may wish
to increase the time between local checkpoints.
Given 5 minutes (300 seconds) per local checkpoint, this means that we need to support writing log records at maximum speed for 6 * 300 = 1800 seconds. The size of a REDO log record is 72 bytes plus 4 bytes per updated column value plus the maximum size of the updated column, and there is one REDO log record for each table record updated in a transaction, on each node where the data reside. Using the numbers of operations set out previously in this section, we derive the following:
50000 select operations per hour yields 0 log records (and
thus 0 bytes), since SELECT
statements
are not recorded in the REDO log.
15000 DELETE
statements per hour is
approximately 5 delete operations per second. (Since we wish
to be conservative in our estimate, we round up here and in
the following calculations.) No columns are updated by
deletes, so these statements consume only 5 operations * 72
bytes per operation = 360 bytes per second.
15000 UPDATE
statements per hour is
roughly the same as 5 updates per second. Each update uses
72 bytes, plus 4 bytes per column * 5 columns updated, plus
32 bytes per column * 5 columns 鈥 this works out to 72
+ 20 + 160 = 252 bytes per operation, and multiplying this
by 5 operation per second yields 1260 bytes per second.
15000 INSERT
statements per hour is
equivalent to 5 insert operations per second. Each insert
requires REDO log space of 72 bytes, plus 4 bytes per record
* 40 columns, plus 32 bytes per column * 40 columns, which
is 72 + 160 + 1280 = 1512 bytes per operation. This times 5
operations per second yields 7560 bytes per second.
So the total number of REDO log bytes being written per second
is approximately 0 + 360 + 1260 + 7560 = 9180 bytes. Mutiplied
by 1800 seconds, this yields 16524000 bytes required for REDO
logging, or approximately 15.75 MB. The unit used for
NoOfFragmentLogFiles
represents a set of 4
16-MB log files 鈥 that is, 64 MB. Thus, the minimum value
(3) for this parameter is sufficient for the scenario envisioned
in this example, since 3 times 64 = 192 MB, or about 12 times
what is required; the default value of 8 (or 512 MB) is more
than ample in this case.
This portion of the MySQL Cluster chapter covers upgrading and downgrading a MySQL Cluster from one MySQL release to another. It discusses different types of Cluster upgrades and downgrades, and provides a Cluster upgrade/downgrade compatibility matrix (see Section聽15.5.2, 鈥淐luster Upgrade and Downgrade Compatibility鈥). You are expected already to be familiar with installing and configuring a MySQL Cluster prior to attempting an upgrade or downgrade. See Section聽15.4, 鈥淢ySQL Cluster Configuration鈥.
This section remains in development, and continues to be updated and expanded.
This section discusses how to perform a rolling restart of a MySQL Cluster installation, so called because it involves stopping and starting (or restarting) each node in turn, so that the cluster itself remains operational. This is often done as part of a rolling upgrade or rolling downgrade, where high availability of the cluster is mandatory and no downtime of the cluster as a whole is permissible. Where we refer to upgrades, the information provided here also generally applies to downgrades as well.
There are a number of reasons why a rolling restart might be desirable:
Cluster Configuration Change: To make a change in the cluster's configuration, such as adding an SQL node to the cluster, or setting a configuration parameter to a new value.
Cluster Software Upgrade/Downgrade: To upgrade the cluster to a newer version of the MySQL Cluster software (or to downgrade it to an older version). This is usually referred to as a 鈥rolling upgrade鈥 (or 鈥rolling downgrade鈥, when reverting to an older version of MySQL Cluster).
Change on Node Host: To make changes in the hardware or operating system on which one or more cluster nodes are running
Cluster Reset: To reset the cluster because it has reached an undesirable state
Freeing of Resources: To
allow memory allocated to a table by successive
INSERT
and DELETE
operations to be freed for re-use by other Cluster tables
The process for performing a rolling restart may be generalised as follows:
Stop all cluster management nodes (ndb_mgmd processes), reconfigure them, then restart them
Stop, reconfigure, then restart each cluster data node (ndbd process) in turn
Stop, reconfigure, then restart each cluster SQL node (mysqld process) in turn
The specifics for implementing a particular rolling upgrade depend upon the actual changes being made. A more detailed view of the process is presented here:
In the previous diagram, Stop
and Start steps indicate that
the process must be stopped completely using a shell command
(such as kill on most Unix systems) or the
management client STOP
command, then started
again from a system shell by invoking the
ndbd or ndb_mgmd
executable as appropriate.
Restart indicates the process
may be restarted using the ndb_mgm management
client RESTART
command.
When performing an upgrade or downgrade of the cluster software, you must upgrade or downgrade the management nodes first, then the data nodes, and finally the SQL nodes. Doing so in any other order may leave the cluster in an unusable state.
This section provides information regarding Cluster software and table file compatibility between differing versions of the MySQL Server for purposes of performing upgrades and downgrades.
Important: Only compatibility
between MySQL versions with regard to NDB
Cluster
is taken into account in this section, and
there are likely other issues to be considered. As
with any other MySQL software upgrade or downgrade, you are
strongly encouraged to review the relevant portions of the MySQL
Manual for the MySQL versions from which and to which you intend
to migrate, before attempting an upgrade or downgrade of the
MySQL Cluster software. See
Section聽2.11, 鈥淯pgrading MySQL鈥.
The following table shows Cluster upgrade and downgrade compatibility between different versions of the MySQL Server.
Notes:
4.1 Series:
You cannot upgrade directly from 4.1.8 to 4.1.10 (or newer); you must first upgrade from 4.1.8 to 4.1.9, then upgrade to 4.1.10. Similarly, you cannot downgrade directly from 4.1.10 (or newer) to 4.1.8; you must first downgrade from 4.1.10 to 4.1.9, then downgrade from 4.1.9 to 4.1.8.
If you wish to upgrade a MySQL Cluster to 4.1.15, you must upgrade to 4.1.14 first, and you must upgrade to 4.1.15 before upgrading to 4.1.16 or newer.
Cluster downgrades from 4.1.15 to 4.1.14 (or earlier versions) are not supported.
Cluster upgrades from MySQL Server versions previous to
4.1.8 are not supported; when upgrading from these, you must
dump all NDB
tables, install the new
version of the software, and then reload the tables from the
dump.
5.0 Series:
MySQL 5.0.2 was the first public release in this series.
Cluster downgrades from MySQL 5.0 to MySQL 4.1 are not supported.
Cluster downgrades from 5.0.12 to 5.0.11 (or earlier) are not supported.
You cannot restore with ndb_restore to a MySQL 5.0 Cluster using a backup made from a Cluster running MySQL 5.1. You must use mysqldump in such cases.
There was no public release for MySQL 5.0.23.
5.1 Series:
MySQL 5.1.3 was the first public release in this series.
You cannot downgrade a MySQL 5.1.6 or later Cluster using Disk Data tables to MySQL 5.1.5 or earlier unless you convert all such tables to in-memory Cluster tables first.
MySQL 5.1.8, MySQL 5.1.10, and MySQL 5.1.13 were not released.
Online cluster upgrades and downgrades between MySQL 5.1.11
(or an earlier version) and 5.1.12 (or a later version) are
not possible due to major changes in the cluster filesystem.
In such cases, you must perform a backup or dump, upgrade
(or downgrade) the software, start each data node with
--initial
, and then restore from the backup
or dump. You can use NDB
backup/restore
or mysqldump for this purpose.
Online downgrades from MySQL 5.1.14 or later to versions previous to 5.1.14 are not supported due to incompatible changes in the cluster system tables.
Online upgrades from MySQL 5.1.17 and earlier to 5.1.18 and
later are not supported for clusters using replication due
to incompatible changes in the
mysql.ndb_apply_status
table. However, it
should not be necessary to shut down the cluster entirely,
if you follow this modified rolling restart procedure:
Stop the management server, update the
ndb_mgmd
binary, then start it
again. For multiple management servers, repeat this
step for each management server in turn.
For each data node in turn: Stop the data node,
replace the ndbd
binary with the
new version, then restart the data node. It is not
necessary to use --initial
when
restarting any of the data nodes.
Stop all SQL nodes. Replace the
mysqld binary with the new version
for all SQL nodes, then restart them. It is not
necessary to start them one at a time, but they must
all be shut down at the same time before starting any
of them again using the 5.1.18 (or later)
mysqld. Otherwise 鈥 due to
the fact that
mysql.ndb_apply_status
uses the
NDB
storage engine and is thus
shared between all SQL nodes 鈥 there may be
conflicts between MySQL servers using the old and new
versions of the table.
You can find more information about the changes to
ndb_apply_status
in
Section聽15.10.4, 鈥淐luster Replication Schema and Tables鈥.
Understanding how to manage MySQL Cluster requires a knowledge of four essential processes. In the next few sections of this chapter, we cover the roles played by these processes in a cluster, how to use them, and what startup options are available for each of them:
mysqld is the traditional MySQL server
process. To be used with MySQL Cluster,
mysqld needs to be built with support for the
NDB Cluster
storage engine, as it is in the
precompiled binaries available from
http://dev.mysql.com/downloads/. If you build MySQL from
source, you must invoke configure with the
--with-ndbcluster
option to enable NDB
Cluster
storage engine support.
If the mysqld binary has been built with
Cluster support, the NDB Cluster
storage
engine is still disabled by default. You can use either of two
possible options to enable this engine:
Use --ndbcluster
as a startup option on the
command line when starting mysqld.
Insert a line containing ndbcluster
in
the [mysqld]
section of your
my.cnf
file.
An easy way to verify that your server is running with the
NDB Cluster
storage engine enabled is to
issue the SHOW ENGINES
statement in the MySQL
Monitor (mysql). You should see the value
YES
as the Support
value
in the row for NDBCLUSTER
. If you see
NO
in this row or if there is no such row
displayed in the output, you are not running an
NDB
-enabled version of MySQL. If you see
DISABLED
in this row, you need to enable it
in either one of the two ways just described.
To read cluster configuration data, the MySQL server requires at a minimum three pieces of information:
The MySQL server's own cluster node ID
The hostname or IP address for the management server (MGM node)
The number of the TCP/IP port on which it can connect to the management server
Node IDs can be allocated dynamically, so it is not strictly necessary to specify them explicitly.
The mysqld parameter
ndb-connectstring
is used to specify the
connectstring either on the command line when starting
mysqld or in my.cnf
. The
connectstring contains the hostname or IP address where the
management server can be found, as well as the TCP/IP port it
uses.
In the following example, ndb_mgmd.mysql.com
is the host where the management server resides, and the
management server listens for cluster messages on port 1186:
shell> mysqld --ndbcluster --ndb-connectstring=ndb_mgmd.mysql.com:1186
See Section聽15.4.4.2, 鈥淭he Cluster Connectstring鈥, for more information on connectstrings.
Given this information, the MySQL server will be a full participant in the cluster. (We often refer to a mysqld process running in this manner as an SQL node.) It will be fully aware of all cluster data nodes as well as their status, and will establish connections to all data nodes. In this case, it is able to use any data node as a transaction coordinator and to read and update node data.
You can see in the mysql client whether a
MySQL server is connected to the cluster using SHOW
PROCESSLIST
. If the MySQL server is connected to the
cluster, and you have the PROCESS
privilege,
then the first row of the output is as shown here:
mysql> SHOW PROCESSLIST \G *************************** 1. row *************************** Id: 1 User: system user Host: db: Command: Daemon Time: 1 State: Waiting for event from ndbcluster Info: NULL
To participate in a MySQL Cluster, the
mysqld process must be started with
both the options
--ndbcluster
and
--ndb-connectstring
(or their equivalents in
my.cnf
). If mysqld is
started with only the --ndbcluster
option, or
if it is unable to contact the cluster, it is not possible to
work with NDB
tables, nor is it
possible to create any new tables regardless of storage
engine. The latter restriction is a safety measure
intended to prevent the creation of tables having the same
names as NDB
tables while the SQL node is
not connected to the cluster. If you wish to create tables
using a different storage engine while the
mysqld process is not participating in a
MySQL Cluster, you must restart the server
without the --ndbcluster
option.
ndbd is the process that is used to handle all the data in tables using the NDB Cluster storage engine. This is the process that empowers a data node to accomplish distributed transaction handling, node recovery, checkpointing to disk, online backup, and related tasks.
In a MySQL Cluster, a set of ndbd processes cooperate in handling data. These processes can execute on the same computer (host) or on different computers. The correspondences between data nodes and Cluster hosts is completely configurable.
ndbd generates a set of log files which
are placed in the directory specified by
DataDir
in the
config.ini
configuration file. These
log files are listed below. Note that
node_id
represents the node's
unique identifier. For example,
ndb_2_error.log
is the error log
generated by the data node whose node ID is
2
.
ndb_
is a file containing records of all crashes which the
referenced ndbd process has
encountered. Each record in this file contains a brief
error string and a reference to a trace file for this
crash. A typical entry in this file might appear as
shown here:
node_id
_error.log
Date/Time: Saturday 30 July 2004 - 00:20:01 Type of error: error Message: Internal program error (failed ndbrequire) Fault ID: 2341 Problem data: DbtupFixAlloc.cpp Object of reference: DBTUP (Line: 173) ProgramName: NDB Kernel ProcessID: 14909 TraceFile: ndb_2_trace.log.2 ***EOM***
Note: It is
very important to be aware that the last entry in the
error log file is not necessarily the newest
one (nor is it likely to be). Entries in the
error log are not listed in
chronological order; rather, they correspond to the
order of the trace files as determined in the
ndb_
file (see below). Error log entries are thus overwritten
in a cyclical and not sequential fashion.
node_id
_trace.log.next
ndb_
is a trace file describing exactly what happened just
before the error occurred. This information is useful
for analysis by the MySQL Cluster development team.
node_id
_trace.log.trace_id
It is possible to configure the number of these trace
files that will be created before old files are
overwritten. trace_id
is a
number which is incremented for each successive trace
file.
ndb_
is the file that keeps track of the next trace file
number to be assigned.
node_id
_trace.log.next
ndb_
is a file containing any data output by the
ndbd process. This file is created
only if ndbd is started as a daemon,
which is the default behavior.
node_id
_out.log
ndb_
is a file containing the process ID of the
ndbd process when started as a
daemon. It also functions as a lock file to avoid the
starting of nodes with the same identifier.
node_id
.pid
ndb_
is a file used only in debug versions of
ndbd, where it is possible to trace
all incoming, outgoing, and internal messages with their
data in the ndbd process.
node_id
_signal.log
It is recommended not to use a directory mounted through NFS
because in some environments this can cause problems whereby
the lock on the .pid
file remains in
effect even after the process has terminated.
To start ndbd, it may also be necessary to specify the hostname of the management server and the port on which it is listening. Optionally, one may also specify the node ID that the process is to use.
shell> ndbd --connect-string="nodeid=2;host=ndb_mgmd.mysql.com:1186"
See Section聽15.4.4.2, 鈥淭he Cluster Connectstring鈥, for additional information about this issue. Section聽15.6.5, 鈥淐ommand Options for MySQL Cluster Processes鈥, describes other options for ndbd.
When ndbd starts, it actually initiates two processes. The first of these is called the 鈥angel process鈥; its only job is to discover when the execution process has been completed, and then to restart the ndbd process if it is configured to do so. Thus, if you attempt to kill ndbd via the Unix kill command, it is necessary to kill both processes, beginning with the angel process. The preferred method of terminating an ndbd process is to use the management client and stop the process from there.
The execution process uses one thread for reading, writing, and scanning data, as well as all other activities. This thread is implemented asynchronously so that it can easily handle thousands of concurrent activites. In addition, a watch-dog thread supervises the execution thread to make sure that it does not hang in an endless loop. A pool of threads handles file I/O, with each thread able to handle one open file. Threads can also be used for transporter connections by the transporters in the ndbd process. In a multi-processor system performing a large number of operations (including updates), the ndbd process can consume up to 2 CPUs if permitted to do so.
For a machine with many CPUs it is possible to use several ndbd processes which belong to different node groups; however, such a configuration is still considered experimental and is not supported for MySQL 5.1 in a production setting. See Section聽15.13, 鈥淜nown Limitations of MySQL Cluster鈥.
The management server is the process that reads the cluster configuration file and distributes this information to all nodes in the cluster that request it. It also maintains a log of cluster activities. Management clients can connect to the management server and check the cluster's status.
It is not strictly necessary to specify a connectstring when starting the management server. However, if you are using more than one management server, a connectstring should be provided and each node in the cluster should specify its node ID explicitly.
See Section聽15.4.4.2, 鈥淭he Cluster Connectstring鈥, for information about using connectstrings. Section聽15.6.5, 鈥淐ommand Options for MySQL Cluster Processes鈥, describes other options for ndb_mgmd.
The following files are created or used by
ndb_mgmd in its starting directory, and
are placed in the DataDir
as specified in
the config.ini
configuration file. In
the list that follows, node_id
is
the unique node identifier.
config.ini
is the configuration
file for the cluster as a whole. This file is created by
the user and read by the management server.
Section聽15.4, 鈥淢ySQL Cluster Configuration鈥, discusses
how to set up this file.
ndb_
is the cluster events log file. Examples of such events
include checkpoint startup and completion, node startup
events, node failures, and levels of memory usage. A
complete listing of cluster events with descriptions may
be found in Section聽15.7, 鈥淢anagement of MySQL Cluster鈥.
node_id
_cluster.log
When the size of the cluster log reaches one million
bytes, the file is renamed to
ndb_
,
where node_id
_cluster.log.seq_id
seq_id
is the sequence
number of the cluster log file. (For example: If files
with the sequence numbers 1, 2, and 3 already exist, the
next log file is named using the number
4
.)
ndb_
is the file used for node_id
_out.logstdout
and
stderr
when running the management
server as a daemon.
ndb_
is the process ID file used when running the management
server as a daemon.
node_id
.pid
The ndb_mgm management client process is actually not needed to run the cluster. Its value lies in providing a set of commands for checking the cluster's status, starting backups, and performing other administrative functions. The management client accesses the management server using a C API. Advanced users can also employ this API for programming dedicated management processes to perform tasks similar to those performed by ndb_mgm.
To start the management client, it is necessary to supply the hostname and port number of the management server:
shell> ndb_mgm [host_name
[port_num
]]
For example:
shell> ndb_mgm ndb_mgmd.mysql.com 1186
The default hostname and port number are
localhost
and 1186, respectively.
Additional information about using ndb_mgm can be found in Section聽15.6.5.4, 鈥淐ommand Options for ndb_mgm鈥, and Section聽15.7.2, 鈥淐ommands in the MySQL Cluster Management Client鈥.
All MySQL Cluster executables (except for
mysqld) take the options described in this
section. Users of earlier MySQL Cluster versions should note
that some of these options have been changed to make them
consistent with one another as well as with
mysqld. You can use the
--help
option with any MySQL Cluster executable
to view a list of the options which it supports.
The following options are common to all MySQL Cluster executables:
--help
--usage
,
-?
Prints a short list with descriptions of the available command options.
--connect-string=
,
connect_string
-c
connect_string
connect_string
sets the
connectstring to the management server as a command option.
shell> ndbd --connect-string="nodeid=2;host=ndb_mgmd.mysql.com:1186"
--debug[=
options
]
This option can be used only for versions compiled with debugging enabled. It is used to enable output from debug calls in the same manner as for the mysqld process.
--execute=
,
command
-e
command
Can be used to send a command to a Cluster executable from the system shell. For example, either of the following:
shell> ndb_mgm -e "SHOW"
or
shell> ndb_mgm --execute="SHOW"
is equivalent to
ndb_mgm> SHOW
This is analogous to how the --execute
or
-e
option works with the
mysql command-line client. See
Section聽4.3.1, 鈥淯sing Options on the Command Line鈥.
--version
, -V
Prints the MySQL Cluster version number of the executable. The version number is relevant because not all versions can be used together, and the MySQL Cluster startup process verifies that the versions of the binaries being used can co-exist in the same cluster. This is also important when performing an online (rolling) software upgrade or downgrade of MySQL Cluster. (See Section聽15.5.1, 鈥淧erforming a Rolling Restart of the Cluster鈥).
The next few sections describe options specific to individual
NDB
programs.
--ndb-connectstring=
connect_string
When using the NDB Cluster
storage
engine, this option specifies the management server that
distributes cluster configuration data.
--ndbcluster
The NDB Cluster
storage engine is
necessary for using MySQL Cluster. If a
mysqld binary includes support for the
NDB Cluster
storage engine, the engine
is disabled by default. Use the
--ndbcluster
option to enable it. Use
--skip-ndbcluster
to explicitly disable
the engine.
For options common to all NDB
programs, see
Section聽15.6.5, 鈥淐ommand Options for MySQL Cluster Processes鈥.
--bind-address
Causes ndbd to bind to a specific network interface. This option has no default value.
This option was added in MySQL 5.1.12.
--daemon
, -d
Instructs ndbd to execute as a daemon
process. This is the default behavior.
--nodaemon
can be used to prevent the
process from running as a daemon.
--initial
Instructs ndbd to perform an initial start. An initial start erases any files created for recovery purposes by earlier instances of ndbd. It also re-creates recovery log files. Note that on some operating systems this process can take a substantial amount of time.
An --initial
start is to be used only the
very first time that the ndbd process
is started because it removes all files from the Cluster
filesystem and re-creates all REDO log files. The
exceptions to this rule are:
When performing a software upgrade which has changed the contents of any files.
When restarting the node with a new version of ndbd.
As a measure of last resort when for some reason the node restart or system restart repeatedly fails. In this case, be aware that this node can no longer be used to restore data due to the destruction of the data files.
This option does not affect either of the following:
Backup files that have already been created by the affected node
Cluster Disk Data files (see Section聽15.11, 鈥淢ySQL Cluster Disk Data Tables鈥).
It is possible to achieve the same effect by deleting by
other means (such as using rm -r -f)
all files and directories in the data node's
DataDir
鈥 with the possible
exception of the BACKUP
directory in
DataDir
, should you wish to retain any
backups that have been created on that data node 鈥
and then starting ndbd without having
to use the --initial
option. This may be
useful when scripting Cluster administrative tasks.
--initial-start
This option is used when performing a partial initial
start of the cluster. Each node should be started with
this option, as well as --no-wait-nodes
.
For example, suppose you have a 4-node cluster whose data nodes have the IDs 2, 3, 4, and 5, and you wish to perform a partial initial start using only nodes 2, 4, and 5 鈥 that is, omitting node 3:
ndbd --ndbd-nodeid=2 --no-wait-nodes=3 --initial-start ndbd --ndbd-nodeid=4 --no-wait-nodes=3 --initial-start ndbd --ndbd-nodeid=5 --no-wait-nodes=3 --initial-start
This option was added in MySQL 5.1.9.
--nowait-nodes=
node_id_1
[,
node_id_2
[, ...]]
This option takes a list of data nodes which for which the cluster will not wait for before starting.
This can be used to start the cluster in a partitioned
state. For example, to start the cluster with only half of
the data nodes (nodes 2, 3, 4, and 5) running in a 4-node
cluster, you can start each ndbd
process with --nowait-nodes=3,5
. In this
case, the cluster starts as soon as nodes 2 and 4 connect,
and does not wait
StartPartitionedTimeout
milliseconds
for nodes 3 and 5 to connect as it would otherwise.
If you wanted to start up the same cluster as in the
previous example without one ndbd
鈥 say, for example, that the host machine for node 3
has suffered a hardware failure 鈥 then start nodes
2, 4, and 5 with --no-wait-nodes=3
. Then
the cluster will start as soon as nodes 2, 4, and 5
connect and will not wait for node 3 to start.
This option was added in MySQL 5.1.9.
--nodaemon
Instructs ndbd not to start as a daemon process. This is useful when ndbd is being debugged and you want output to be redirected to the screen.
--nostart
, -n
Instructs ndbd not to start
automatically. When this option is used,
ndbd connects to the management server,
obtains configuration data from it, and initializes
communication objects. However, it does not actually start
the execution engine until specifically requested to do so
by the management server. This can be accomplished by
issuing the proper START
command in the
management client (see
Section聽15.7.2, 鈥淐ommands in the MySQL Cluster Management Client鈥).
For options common to NDB programs, see Section聽15.6.5, 鈥淐ommand Options for MySQL Cluster Processes鈥.
--config-file=
,
filename
-f
filename
Instructs the management server as to which file it should
use for its configuration file. This option must be
specified. The filename defaults to
config.ini
.
Note: This option also
can be given as -c
, but this
shortcut is obsolete and should not
be used in new installations.
file_name
--daemon
, -d
Instructs ndb_mgmd to start as a daemon process. This is the default behavior.
--nodaemon
Instructs ndb_mgmd not to start as a daemon process.
For options common to NDB programs, see Section聽15.6.5, 鈥淐ommand Options for MySQL Cluster Processes鈥.
--try-reconnect=
number
If the connection to the management server is broken, the
node tries to reconnect to it every 5 seconds until it
succeeds. By using this option, it is possible to limit
the number of attempts to
number
before giving up and
reporting an error instead.
Managing a MySQL Cluster involves a number of tasks, the first of which is to configure and start MySQL Cluster. This is covered in Section聽15.4, 鈥淢ySQL Cluster Configuration鈥, and Section聽15.6, 鈥淧rocess Management in MySQL Cluster鈥.
The following sections cover the management of a running MySQL Cluster.
There are essentially two methods of actively managing a running
MySQL Cluster. The first of these is through the use of commands
entered into the management client whereby cluster status can be
checked, log levels changed, backups started and stopped, and
nodes stopped and started. The second method involves studying the
contents of the cluster log
ndb_
;
this is usually found in the management server's
node_id
_cluster.logDataDir
directory, but this location can be
overridden using the LogDestination
option
鈥 see Section聽15.4.4.4, 鈥淒efining the Management Server鈥, for
details. (Recall that node_id
represents the unique identifier of the node whose activity is
being logged.) The cluster log contains event reports generated by
ndbd. It is also possible to send cluster log
entries to a Unix system log.
This section describes the steps involved when the cluster is started.
There are several different startup types and modes, as shown here:
Initial Start: The cluster
starts with a clean filesystem on all data nodes. This
occurs either when the cluster started for the very first
time, or when it is restarted using the
--initial
option.
Note: Disk Data files are not removed
when restarting a node using --initial
.
System Restart: The cluster starts and reads data stored in the data nodes. This occurs when the cluster has been shut down after having been in use, when it is desired for the cluster to resume operations from the point where it left off.
Node Restart: This is the online restart of a cluster node while the cluster itself is running.
Initial Node Restart: This is the same as a node restart, except that the node is reinitialized and started with a clean filesystem.
Prior to startup, each data node (ndbd
process) must be initialized. Initialization consists of the
following steps:
Obtain a Node ID.
Fetch configuration data.
Allocate ports to be used for inter-node communications.
Allocate memory according to settings obtained from the configuration file.
When a data node or SQL node first connects to the management node, it reserves a cluster node ID. To make sure that no other node allocates the same node ID, this ID is retained until the node has managed to connect to the cluster and at least one ndbd reports that this node is connected. This retention of the node ID is guarded by the connection between the node in question and ndb_mgmd.
Normally, in the event of a problem with the node, the node disconnects from the management server, the socket used for the connection is closed, and the reserved node ID is freed. However, if a node is disconnected abruptly 鈥 for example, due to a hardware failure in one of the cluster hosts, or because of network issues 鈥 the normal closing of the socket by the operating system may not take place. In this case, the node ID continues to be reserved and not released until a TCP timeout occurs 10 or so minutes later.
To take care of this problem, you can use PURGE STALE
SESSIONS
. Running this statement forces all reserved
node IDs to be checked; any that are not being used by nodes
actually connected to the cluster are then freed.
Beginning with MySQL 5.1.11, timeout handling of node ID
assignments is implemented. This performs the ID usage checks
automatically after approximately 20 seconds, so that
PURGE STALE SESSIONS
should no longer be
necessary in a normal Cluster start.
After each data node has been initialized, the cluster startup process can proceed. The stages which the cluster goes through during this process are listed here:
Stage 0
Clear the cluster filesystem. This stage occurs
only if the cluster was started with
the --initial
option.
Stage 1
This stage sets up Cluster connections, establishes inter-node communications, and starts Cluster heartbeats.
Note: When one or more nodes hang in Phase 1 while the remaining node or nodes hang in Phase 2, this often indicates network problems. One possible cause of such issues is one or more cluster hosts having multiple network interfaces. Another common source of problems causing this condition is the blocking of TCP/IP ports needed for communications between cluster nodes. In the latter case, this is often due to a misconfigured firewall.
Stage 2
The arbitrator node is elected. If this is a system restart, the cluster determines the latest restorable global checkpoint.
Stage 3
This stage initializes a number of internal cluster variables.
Stage 4
For an initial start or initial node restart, the redo log
files are created. The number of these files is equal to
NoOfFragmentLogFiles
.
For a system restart:
Read schema or schemas.
Read data from the local checkpoint.
Apply all redo information until the latest restorable global checkpoint has been reached.
For a node restart, find the tail of the redo log.
Stage 5
If this is an initial start, create the
SYSTAB_0
and
NDB$EVENTS
internal system tables.
For a node restart or an initial node restart:
The node is included in transaction handling operations.
The node schema is compared with that of the master and synchronized with it.
Synchronize data received in the form of
INSERT
from the other data nodes in
this node's node group.
In all cases, wait for complete local checkpoint as determined by the arbitrator.
Stage 6
Update internal variables.
Stage 7
Update internal variables.
Stage 8
In a system restart, rebuild all indexes.
Stage 9
Update internal variables.
Stage 100
At this point in a node restart or initial node restart, APIs may connect to the node and begin to receive events.
Stage 101
At this point in a node restart or initial node restart, event delivery is handed over to the node joining the cluster. The newly-joined node takes over responsibility for delivering its primary data to subscribers.
After this process is completed for an initial start or system restart, transaction handling is enabled. For a node restart or initial node restart, completion of the startup process means that the node may now act as a transaction coordinator.
In addition to the central configuration file, a cluster may also be controlled through a command-line interface available through the management client ndb_mgm. This is the primary administrative interface to a running cluster.
Commands for the event logs are given in Section聽15.7.3, 鈥淓vent Reports Generated in MySQL Cluster鈥; commands for creating backups and restoring from backup are provided in Section聽15.8, 鈥淥n-line Backup of MySQL Cluster鈥.
The management client has the following basic commands. In the
listing that follows, node_id
denotes
either a database node ID or the keyword ALL
,
which indicates that the command should be applied to all of the
cluster's data nodes.
HELP
Displays information on all available commands.
SHOW
Displays information on the cluster's status.
Note: In a cluster where multiple management nodes are in use, this command displays information only for data nodes that are actually connected to the current management server.
node_id
START
Brings online the data node identified by
node_id
(or all data nodes).
ALL START
works on all data nodes only,
and does not affect management nodes.
Important: To use this command to bring a data node online, the data node must have been started using ndbd --nostart or ndbd -n.
node_id
STOP
Stops the data or management node identified by
node_id
. Note that ALL
STOP
works to stop all data nodes only, and does
not affect management nodes.
A node affected by this command disconnects from the cluster, and its associated ndbd or ndb_mgmd process terminates.
node_id
RESTART [-n]
[-i]
Restarts the data node identified by
node_id
(or all data nodes).
Using the -i
option with
RESTART
causes the data node to perform
an initial restart; that is, the node's filesystem is
deleted and recreated. The effect is the same as that
obtained from stopping the data node process and then
starting it again using ndbd --initial
from the system shell. Note that backup files and Disk Data
files are not removed when this option is used.
Using the -n
option causes the data node
process to be restarted, but the data node is not actually
brought online until the appropriate
START
command is issued. The effect of
this option is the same as that obtained from stopping the
data node and then starting it again using ndbd
--nostart
or ndbd -n
from the
system shell.
node_id
STATUS
Displays status information for the data node identified by
node_id
(or for all data nodes).
ENTER SINGLE USER MODE
node_id
Enters single user mode, whereby only the MySQL server
identified by the node ID node_id
is allowed to access the database.
It is not possible in MySQL 5.1 for data nodes to join the cluster while it is running in single user mode. (See Bug#20395 for more information.)
EXIT SINGLE USER MODE
Exits single user mode, allowing all SQL nodes (that is, all running mysqld processes) to access the database.
QUIT
, EXIT
Terminates the management client.
This command does not affect any nodes connected to the cluster.
SHUTDOWN
Shuts down all cluster data nodes and management nodes. To
exit the management client after this has been done, use
EXIT
or QUIT
.
This command does not shut down any SQL nodes or API nodes that are connected to the cluster.
In this section, we discuss the types of event logs provided by MySQL Cluster, and the types of events that are logged.
MySQL Cluster provides two types of event log:
The cluster log, which includes events generated by all cluster nodes. The cluster log is the log recommended for most uses because it provides logging information for an entire cluster in a single location.
By default, the cluster log is saved to a file named
ndb_
,
(where node_id
_cluster.lognode_id
is the node ID of
the management server) in the same directory where the
ndb_mgm binary resides.
Cluster logging information can also be sent to
stdout
or a syslog
facility in addition to or instead of being saved to a file,
as determined by the values set for the
DataDir
and
LogDestination
configuration parameters.
See Section聽15.4.4.4, 鈥淒efining the Management Server鈥, for more
information about these parameters.
Node logs are local to each node.
Output generated by node event logging is written to the
file
ndb_
(where node_id
_out.lognode_id
is the node's node
ID) in the node's DataDir
. Node event
logs are generated for both management nodes and data nodes.
Node logs are intended to be used only during application development, or for debugging application code.
Both types of event logs can be set to log different subsets of events.
Each reportable event can be distinguished according to three different criteria:
Category: This can be any one of the
following values: STARTUP
,
SHUTDOWN
, STATISTICS
,
CHECKPOINT
,
NODERESTART
,
CONNECTION
, ERROR
, or
INFO
.
Priority: This is represented by one of the numbers from 1 to 15 inclusive, where 1 indicates 鈥most important鈥 and 15 鈥least important.鈥
Severity Level: This can be any one of
the following values: ALERT
,
CRITICAL
, ERROR
,
WARNING
, INFO
, or
DEBUG
.
Both the cluster log and the node log can be filtered on these properties.
The format used in the cluster log is as shown here:
2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 1: Data usage is 2%(60 32K pages of total 2560) 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 1: Index usage is 1%(24 8K pages of total 2336) 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 1: Resource 0 min: 0 max: 639 curr: 0 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 2: Data usage is 2%(76 32K pages of total 2560) 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 2: Index usage is 1%(24 8K pages of total 2336) 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 2: Resource 0 min: 0 max: 639 curr: 0 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 3: Data usage is 2%(58 32K pages of total 2560) 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 3: Index usage is 1%(25 8K pages of total 2336) 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 3: Resource 0 min: 0 max: 639 curr: 0 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 4: Data usage is 2%(74 32K pages of total 2560) 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 4: Index usage is 1%(25 8K pages of total 2336) 2007-01-26 19:35:55 [MgmSrvr] INFO -- Node 4: Resource 0 min: 0 max: 639 curr: 0 2007-01-26 19:39:42 [MgmSrvr] INFO -- Node 4: Node 9 Connected 2007-01-26 19:39:42 [MgmSrvr] INFO -- Node 1: Node 9 Connected 2007-01-26 19:39:42 [MgmSrvr] INFO -- Node 1: Node 9: API version 5.1.15 2007-01-26 19:39:42 [MgmSrvr] INFO -- Node 2: Node 9 Connected 2007-01-26 19:39:42 [MgmSrvr] INFO -- Node 2: Node 9: API version 5.1.15 2007-01-26 19:39:42 [MgmSrvr] INFO -- Node 3: Node 9 Connected 2007-01-26 19:39:42 [MgmSrvr] INFO -- Node 3: Node 9: API version 5.1.15 2007-01-26 19:39:42 [MgmSrvr] INFO -- Node 4: Node 9: API version 5.1.15 2007-01-26 19:59:22 [MgmSrvr] ALERT -- Node 2: Node 7 Disconnected 2007-01-26 19:59:22 [MgmSrvr] ALERT -- Node 2: Node 7 Disconnected
Each line in the cluster log contains the following information:
A timestamp in
format.
YYYY
-MM
-DD
HH
:MM
:SS
The type of node which is performing the logging. In the
cluster log, this is always [MgmSrvr]
.
The severity of the event.
The ID of the node reporting the event.
A description of the event. The most common types of events to appear in the log are connections and disconnections between different nodes in the cluster, and when checkpoints occur. In some cases, the description may contain status information.
The following management commands are related to the cluster log:
CLUSTERLOG ON
Turns the cluster log on.
CLUSTERLOG OFF
Turns the cluster log off.
CLUSTERLOG INFO
Provides information about cluster log settings.
node_id
CLUSTERLOG
category
=threshold
Logs category
events with
priority less than or equal to
threshold
in the cluster log.
CLUSTERLOG FILTER
severity_level
Toggles cluster logging of events of the specified
severity_level
.
The following table describes the default setting (for all data nodes) of the cluster log category threshold. If an event has a priority with a value lower than or equal to the priority threshold, it is reported in the cluster log.
Note that events are reported per data node, and that the threshold can be set to different values on different nodes.
Category | Default threshold (All data nodes) |
STARTUP | 7 |
SHUTDOWN | 7 |
STATISTICS | 7 |
CHECKPOINT | 7 |
NODERESTART | 7 |
CONNECTION | 7 |
ERROR | 15 |
INFO | 7 |
The STATISTICS
category can provide a great
deal of useful data. See
Section聽15.7.3.3, 鈥淯sing CLUSTERLOG STATISTICS
鈥, for more
information.
Thresholds are used to filter events within each category. For
example, a STARTUP
event with a priority of
3 is not logged unless the threshold for
STARTUP
is changed to 3 or lower. Only
events with priority 3 or lower are sent if the threshold is
3.
The following table shows the event severity levels.
(Note: These correspond to
Unix syslog
levels, except for
LOG_EMERG
and
LOG_NOTICE
, which are not used or mapped.)
1 | ALERT | A condition that should be corrected immediately, such as a corrupted system database |
2 | CRITICAL | Critical conditions, such as device errors or insufficient resources |
3 | ERROR | Conditions that should be corrected, such as configuration errors |
4 | WARNING | Conditions that are not errors, but that might require special handling |
5 | INFO | Informational messages |
6 | DEBUG | Debugging messages used for NDB Cluster development |
Event severity levels can be turned on or off (using
CLUSTERLOG FILTER
鈥 see above). If a
severity level is turned on, then all events with a priority
less than or equal to the category thresholds are logged. If
the severity level is turned off then no events belonging to
that severity level are logged.
An event report reported in the event logs has the following format:
datetime
[string
]severity
--message
For example:
09:19:30 2005-07-24 [NDB] INFO -- Node 4 Start phase 4 completed
This section discusses all reportable events, ordered by category and severity level within each category.
In the event descriptions, GCP and LCP mean 鈥Global Checkpoint鈥 and 鈥Local Checkpoint,鈥 respectively.
CONNECTION
Events
These events are associated with connections between Cluster nodes.
Event | Priority | Severity Level | Description |
data nodes connected | 8 | INFO | Data nodes connected |
data nodes disconnected | 8 | INFO | Data nodes disconnected |
Communication closed | 8 | INFO | SQL node or data node connection closed |
Communication opened | 8 | INFO | SQL node or data node connection opened |
CHECKPOINT
Events
The logging messages shown here are associated with checkpoints.
Event | Priority | Severity Level | Description |
LCP stopped in calc keep GCI | 0 | ALERT | LCP stopped |
Local checkpoint fragment completed | 11 | INFO | LCP on a fragment has been completed |
Global checkpoint completed | 10 | INFO | GCP finished |
Global checkpoint started | 9 | INFO | Start of GCP: REDO log is written to disk |
Local checkpoint completed | 8 | INFO | LCP completed normally |
Local checkpoint started | 7 | INFO | Start of LCP: data written to disk |
STARTUP
Events
The following events are generated in response to the startup of a node or of the cluster and of its success or failure. They also provide information relating to the progress of the startup process, including information concerning logging activities.
Event | Priority | Severity Level | Description |
Internal start signal received STTORRY | 15 | INFO | Blocks received after completion of restart |
New REDO log started | 10 | INFO | GCI keep X , newest restorable GCI
Y |
New log started | 10 | INFO | Log part X , start MB
Y , stop MB
Z |
Node has been refused for inclusion in the cluster | 8 | INFO | Node cannot be included in cluster due to misconfiguration, inability to establish communication, or some other problem |
data node neighbors | 8 | INFO | Shows neighboring data nodes |
data node start phase X completed | 4 | INFO | A data node start phase has been completed |
Node has been successfully included into the cluster | 3 | INFO | Displays the node, managing node, and dynamic ID |
data node start phases initiated | 1 | INFO | NDB Cluster nodes starting |
data node all start phases completed | 1 | INFO | NDB Cluster nodes started |
data node shutdown initiated | 1 | INFO | Shutdown of data node has commenced |
data node shutdown aborted | 1 | INFO | Unable to shut down data node normally |
NODERESTART
Events
The following events are generated when restarting a node and relate to the success or failure of the node restart process.
Event | Priority | Severity Level | Description |
Node failure phase completed | 8 | ALERT | Reports completion of node failure phases |
Node has failed, node state was X | 8 | ALERT | Reports that a node has failed |
Report arbitrator results | 2 | ALERT | There are eight different possible results for arbitration attempts:
|
Completed copying a fragment | 10 | INFO | 聽 |
Completed copying of dictionary information | 8 | INFO | 聽 |
Completed copying distribution information | 8 | INFO | 聽 |
Starting to copy fragments | 8 | INFO | 聽 |
Completed copying all fragments | 8 | INFO | 聽 |
GCP takeover started | 7 | INFO | 聽 |
GCP takeover completed | 7 | INFO | 聽 |
LCP takeover started | 7 | INFO | 聽 |
LCP takeover completed (state = X ) | 7 | INFO | 聽 |
Report whether an arbitrator is found or not | 6 | INFO | There are seven different possible outcomes when seeking an arbitrator:
|
STATISTICS
Events
The following events are of a statistical nature. They provide information such as numbers of transactions and other operations, amount of data sent or received by individual nodes, and memory usage.
Event | Priority | Severity Level | Description |
Report job scheduling statistics | 9 | INFO | Mean internal job scheduling statistics |
Sent number of bytes | 9 | INFO | Mean number of bytes sent to node X |
Received # of bytes | 9 | INFO | Mean number of bytes received from node X |
Report transaction statistics | 8 | INFO | Numbers of: transactions, commits, reads, simple reads, writes, concurrent operations, attribute information, and aborts |
Report operations | 8 | INFO | Number of operations |
Report table create | 7 | INFO | 聽 |
Memory usage | 5 | INFO | Data and index memory usage (80%, 90%, and 100%) |
ERROR
Events
These events relate to Cluster errors and warnings. The presence of one or more of these generally indicates that a major malfunction or failure has occurred.
Event | Priority | Severity | Description |
Dead due to missed heartbeat | 8 | ALERT | Node X declared 鈥dead鈥 due to
missed heartbeat |
Transporter errors | 2 | ERROR | 聽 |
Transporter warnings | 8 | WARNING | 聽 |
Missed heartbeats | 8 | WARNING | Node X missed heartbeat
#Y |
General warning events | 2 | WARNING | 聽 |
INFO
Events
These events provide general information about the state of the cluster and activities associated with Cluster maintenance, such as logging and heartbeat transmission.
Event | Priority | Severity | Description |
Sent heartbeat | 12 | INFO | Heartbeat sent to node X |
Create log bytes | 11 | INFO | Log part, log file, MB |
General information events | 2 | INFO | 聽 |
The NDB
management client's
CLUSTERLOG STATISTICS
command can provide a
number of useful statistics in its output. The following
statistics are reported by the transaction coordinator:
Statistic | Description (Number of...) |
Trans. Count | Transactions attempted with this node as coordinator |
Commit Count | Transactions committed with this node as coordinator |
Read Count | Primary key reads (all) |
Simple Read Count | Primary key reads reading the latest committed value |
Write Count | Primary key writes (includes all INSERT ,
UPDATE , and
DELETE operations) |
AttrInfoCount | Data words used to describe all reads and writes received |
Concurrent Operations | All concurrent operations ongoing at the moment the report is taken |
Abort Count | Transactions with this node as coordinator that were aborted |
Scans | Scans (all) |
Range Scans | Index scans |
The ndbd process has a scheduler that runs in an infinite loop. During each loop scheduler performs the following tasks:
Read any incoming messages from sockets into a job buffer.
Check whether there are any timed messages to be executed; if so, put these into the job buffer as well.
Execute (in a loop) any messages in the job buffer.
Send any distributed messages that were generated by executing the messages in the job buffer.
Wait for any new incoming messages.
The number of loops executed in the third step is reported as
the Mean Loop Counter
. This statistic
increases in size as the utilisation of the TCP/IP buffer
improves. You can use this to monitor performance as you add
new processes to the cluster.
The Mean send size
and Mean
receive size
statistics allow you to gauge the
efficiency of writes and reads (respectively) between nodes.
These values are given in bytes. Higher values mean a lower
cost per byte sent or received; the maximum is 64k.
To cause all cluster log statistics to be logged, you can use
the following command in the NDB
management
client:
ndb_mgm> ALL CLUSTERLOG STATISTICS=15
Note: Setting the threshold
for STATISTICS
to 15 causes the cluster log
to become very verbose, and to gow quite rapidly in size, in
direct proportion to the number of cluster nodes and the
amount of activity on the cluster.
Single user mode allows the database administrator to restrict access to the database system to a single API node, such as a MySQL server (SQL node) or an instance of ndb_restore. When entering single user mode, connections to all other API nodes are closed gracefully and all running transactions are aborted. No new transactions are permitted to start.
Once the cluster has entered single user mode, only the designated API node is granted access to the database.
You can use the ALL STATUS command to see when the cluster has entered single user mode.
Example:
ndb_mgm> ENTER SINGLE USER MODE 5
After this command has executed and the cluster has entered
single user mode, the API node whose node ID is
5
becomes the cluster's only permitted user.
The node specified in the preceding command must be an API node; attempting to specify any other type of node will be rejected.
Note: When the preceding commmand is invoked, all transactions running on the designated node are aborted, the connection is closed, and the server must be restarted.
The command EXIT SINGLE USER MODE changes the state of the cluster's data nodes from single user mode to normal mode. API nodes 鈥 such as MySQL Servers 鈥 waiting for a connection (that is, waiting for the cluster to become ready and available), are again permitted to connect. The API node denoted as the single-user node continues to run (if still connected) during and after the state change.
Example:
ndb_mgm> EXIT SINGLE USER MODE
There are two recommended ways to handle a node failure when running in single user mode:
Method 1:
Finish all single user mode transactions
Issue the EXIT SINGLE USER MODE command
Restart the cluster's data nodes
Method 2:
Restart database nodes prior to entering single user mode.
This section discusses several SQL statements that can prove useful in managing and monitoring a MySQL server that is connected to a MySQL Cluster, and in some cases provide information about the cluster itself.
SHOW ENGINE NDB STATUS
, SHOW
ENGINE NDBCLUSTER STATUS
The output of this statement contains information about the server's connection to the cluster, creation and usage of MySQL Cluster objects, and binary logging for MySQL Cluster replication.
See Section聽13.5.4.12, 鈥SHOW ENGINE
Syntax鈥, for a usage example and
more detailed information.
SHOW ENGINES
This statement can be used to determine whether or not clustering support is enabled in the MySQL server, and if so, whether it is active.
See Section聽13.5.4.13, 鈥SHOW ENGINES
Syntax鈥, for more detailed
information.
In MySQL 5.1, this statement no longer supports a
LIKE
clause. However, you can use
LIKE
to filter queries against the
INFORMATION_SCHEMA.ENGINES
, as
discussed in the next item.
SELECT * FROM INFORMATION_SCHEMA.ENGINES [WHERE
ENGINE LIKE 'NDB%']
This is the equivalent of SHOW ENGINES
,
but uses the ENGINES
table of the
INFORMATION_SCHEMA
database (available
beginning with MySQL 5.1.5). Unlike the case with the
SHOW ENGINES
statement, it is possible to
filter the results using a LIKE
clause,
and to select specific columns to obtain information that
may be of use in scripts. For example, the following query
shows whether the server was built with
NDB
support and, if so, whether it is
enabled:
mysql>SELECT SUPPORT FROM INFORMATION_SCHEMA.ENGINES
->WHERE ENGINE LIKE 'NDB%';
+---------+ | support | +---------+ | ENABLED | +---------+
See Section聽22.18, 鈥淭he INFORMATION_SCHEMA ENGINES
Table鈥, for more information.
SHOW VARIABLES LIKE 'NDB%'
This statement provides a list of most server system
variables relating to the NDB
storage
engine, and their values, as shown here:
mysql> SHOW VARIABLES LIKE 'NDB%';
+-------------------------------------+-------+
| Variable_name | Value |
+-------------------------------------+-------+
| ndb_autoincrement_prefetch_sz | 32 |
| ndb_cache_check_time | 0 |
| ndb_extra_logging | 0 |
| ndb_force_send | ON |
| ndb_index_stat_cache_entries | 32 |
| ndb_index_stat_enable | OFF |
| ndb_index_stat_update_freq | 20 |
| ndb_report_thresh_binlog_epoch_slip | 3 |
| ndb_report_thresh_binlog_mem_usage | 10 |
| ndb_use_copying_alter_table | OFF |
| ndb_use_exact_count | ON |
| ndb_use_transactions | ON |
+-------------------------------------+-------+
See Section聽5.2.3, 鈥淪ystem Variables鈥, for more information.
SELECT * FROM INFORMATION_SCHEMA.GLOBAL_VARIABLES
WHERE VARIABLE_NAME LIKE 'NDB%';
This statement is the equivalent of the
SHOW
described in the previous item, and
provides almost identical output, as shown here:
mysql>SELECT * FROM INFORMATION_SCHEMA.GLOBAL_VARIABLES
->WHERE VARIABLE_NAME LIKE 'NDB%';
+-------------------------------------+----------------+ | VARIABLE_NAME | VARIABLE_VALUE | +-------------------------------------+----------------+ | NDB_AUTOINCREMENT_PREFETCH_SZ | 32 | | NDB_CACHE_CHECK_TIME | 0 | | NDB_EXTRA_LOGGING | 0 | | NDB_FORCE_SEND | ON | | NDB_INDEX_STAT_CACHE_ENTRIES | 32 | | NDB_INDEX_STAT_ENABLE | OFF | | NDB_INDEX_STAT_UPDATE_FREQ | 20 | | NDB_REPORT_THRESH_BINLOG_EPOCH_SLIP | 3 | | NDB_REPORT_THRESH_BINLOG_MEM_USAGE | 10 | | NDB_USE_COPYING_ALTER_TABLE | OFF | | NDB_USE_EXACT_COUNT | ON | | NDB_USE_TRANSACTIONS | ON | +-------------------------------------+----------------+
Unlike the case with the SHOW
statement,
it is possible to select individual columns. For example:
mysql>SELECT VARIABLE_VALUE
->FROM INFORMATION_SCHEMA.GLOBAL_VARIABLES
->WHERE VARIABLE_NAME = 'ndb_force_send';
+----------------+ | VARIABLE_VALUE | +----------------+ | ON | +----------------+
See Section聽22.25, 鈥淭he INFORMATION_SCHEMA GLOBAL_VARIABLES
and
SESSION_VARIABLES
Tables鈥, and
Section聽5.2.3, 鈥淪ystem Variables鈥, for more
information.
SHOW STATUS LIKE 'NDB%'
This statement shows at a glance whether or not the MySQL server is acting as a cluster SQL node, and if so, it provides the MySQL server's cluster node ID, the hostname and port for the cluster management server to which it is connected, and the number of data nodes in the cluster, as shown here:
mysql> SHOW STATUS LIKE 'NDB%';
+--------------------------+---------------+
| Variable_name | Value |
+--------------------------+---------------+
| Ndb_cluster_node_id | 10 |
| Ndb_config_from_host | 192.168.0.103 |
| Ndb_config_from_port | 1186 |
| Ndb_number_of_data_nodes | 4 |
+--------------------------+---------------+
If the MySQL server was built with clustering support, but it is not connected to a cluster, all rows in the output of this statement contain a zero or an empty string:
mysql> SHOW STATUS LIKE 'NDB%';
+--------------------------+-------+
| Variable_name | Value |
+--------------------------+-------+
| Ndb_cluster_node_id | 0 |
| Ndb_config_from_host | |
| Ndb_config_from_port | 0 |
| Ndb_number_of_data_nodes | 0 |
+--------------------------+-------+
See also Section聽13.5.4.26, 鈥SHOW STATUS
Syntax鈥.
SELECT * FROM INFORMATION_SCHEMA.GLOBAL_STATUS
WHERE VARIABLE_NAME LIKE 'NDB%';
Beginning with MySQL 5.1.12, this statement provides similar
output to the SHOW
statement discussed in
the previous item. However, unlike the case with
SHOW STATUS
, it is possible using the
SELECT
to extract values in SQL for use
in scripts for monitoring and automation purposes.
See Section聽22.24, 鈥淭he INFORMATION_SCHEMA GLOBAL_STATUS
and
SESSION_STATUS
Tables鈥, for more information.
This section describes how to create a backup and how to restore the database from a backup at a later time.
A backup is a snapshot of the database at a given time. The backup consists of three main parts:
Metadata: the names and definitions of all database tables
Table records: the data actually stored in the database tables at the time that the backup was made
Transaction log: a sequential record telling how and when data was stored in the database
Each of these parts is saved on all nodes participating in the backup. During backup, each node saves these three parts into three files on disk:
BACKUP-
backup_id
.node_id
.ctl
A control file containing control information and metadata. Each node saves the same table definitions (for all tables in the cluster) to its own version of this file.
BACKUP-
backup_id
-0.node_id
.data
A data file containing the table records, which are saved on a per-fragment basis. That is, different nodes save different fragments during the backup. The file saved by each node starts with a header that states the tables to which the records belong. Following the list of records there is a footer containing a checksum for all records.
BACKUP-
backup_id
.node_id
.log
A log file containing records of committed transactions. Only transactions on tables stored in the backup are stored in the log. Nodes involved in the backup save different records because different nodes host different database fragments.
In the listing above, backup_id
stands for the backup identifier and
node_id
is the unique identifier for
the node creating the file.
Before starting a backup, make sure that the cluster is properly configured for performing one. (See Section聽15.8.4, 鈥淐onfiguration for Cluster Backup鈥.)
Creating a backup using the management client involves the following steps:
Start the management client (ndb_mgm).
Execute the command START BACKUP
.
The management client responds as shown here:
Waiting for completed, this may take several minutes Node 1: Backupbackup_id
started from nodemanagement_node_id
Here, backup_id
is the unique
identifier for this particular backup. (This identifier will
also be saved in the cluster log, if it has not been
configured otherwise.)
management_node_id
is the node ID
of the management to which the management client is
connected.
This means that the cluster has received and processed the backup request. It does not mean that the backup has been completed.
Note: Backup messages were not recorded in the cluster log in MySQL 5.1.12 or 5.1.13. The logging of backup operations was restored in MySQL 5.1.14 (see Bug#24544).
When the backup is completed, the management client will indicate this as shown here:
Node 1: Backupbackup_id
started from nodemanagement_node_id
completed StartGCP: 417599 StopGCP: 417602 #Records: 105957 #LogRecords: 0 Data: 99719356 bytes Log: 0 bytes
The values shown for StartGCP
,
StopGCP
, #Records
,
#LogRecords
, Data
, and
Log
will vary according to the specifics
of your cluster.
Cluster backups are created by default in the
BACKUP
subdirectory of the
DataDir
on each data node. This can be
overridden for one or more data nodes individually, or for all
cluster data nodes in the config.ini
file
using the BackupDataDir
configuration
parameter as discussed in
Identifying
Data Nodes. The backup files created for a backup with a
given backup_id
are stored in a
subdirectory named
BACKUP-
in the backup directory.
backup_id
To abort a backup that is already in progress:
Start the management client.
Execute this command:
ndb_mgm> ABORT BACKUP backup_id
The number backup_id
is the
identifier of the backup that was included in the response
of the management client when the backup was started (in the
message Backup
).
backup_id
started from node
management_node_id
The management client will acknowledge the abort request
with Abort of backup
.
Note: At this point, the
management client has not yet received a response from the
cluster data nodes to this request, and the backup has not
yet actually been aborted.
backup_id
ordered
After the backup has been aborted, the management client will report this fact in a manner similar to what is shown here:
Node 1: Backup 3 started from 5 has been aborted. Error: 1321 - Backup aborted by user request: Permanent error: User defined error Node 3: Backup 3 started from 5 has been aborted. Error: 1323 - 1323: Permanent error: Internal error Node 2: Backup 3 started from 5 has been aborted. Error: 1323 - 1323: Permanent error: Internal error Node 4: Backup 3 started from 5 has been aborted. Error: 1323 - 1323: Permanent error: Internal error
In this example, we have shown sample output for a cluster
with 4 data nodes, where the sequence number of the backup
to be aborted is 3
, and the management
node to which the cluster management client is connected has
the node ID 5
. The first node to complete
its part in aborting the backup reports that the reason for
the abort was due to a request by the user. (The remaining
nodes report that the backup was aborted due to an
unspecified internal error.)
Note: There is no guarantee
that the cluster nodes will respond to an ABORT
BACKUP
command in any particular order.
The Backup
messages mean that the backup has been
terminated and that all files relating to this backup have
been removed from the cluster filesystem.
backup_id
started from node
management_node_id
has been
aborted
It is also possible to abort a backup in progress from a system shell using this command:
shell> ndb_mgm -e "ABORT BACKUP backup_id
"
Note: If there is no backup
with ID backup_id
running when an
ABORT BACKUP
is issued, the management client
makes no response, nor is it indicated in the cluster log that
an invalid abort command was sent.
The cluster restoration program is implemented as a separate
command-line utility ndb_restore, which
can normally be found in the MySQL bin
directory. This program reads the files created as a result
of the backup and inserts the stored information into the
database.
ndb_restore must be executed once for
each of the backup files that were created by the
START BACKUP
command used to create the
backup (see
Section聽15.8.2, 鈥淯sing The Management Client to Create a Backup鈥).
This is equal to the number of data nodes in the cluster at
the time that the backup was created.
Note: Before using ndb_restore, it is recommended that the cluster be running in single user mode, unless you are restoring multiple data nodes in parallel. See Section聽15.7.4, 鈥淪ingle User Mode鈥, for more information about single user mode.
Typical options for this utility are shown here:
ndb_restore [-cconnectstring
] -nnode_id
[-s] [-m] -bbackup_id
-r [backup_path=]/path/to/backup/files
[-e]
The -c
option is used to specify a
connectstring which tells ndb_restore
where to locate the cluster management server. (See
Section聽15.4.4.2, 鈥淭he Cluster Connectstring鈥, for
information on connectstrings.) If this option is not used,
then ndb_restore attempts to connect to a
management server on localhost:1186
. This
utility acts as a cluster API node, and so requires a free
connection 鈥slot鈥 to connect to the cluster
management server. This means that there must be at least
one [API]
or [MYSQLD]
section that can be used by it in the cluster
config.ini
file. It is a good idea to
keep at least one empty [API]
or
[MYSQLD]
section in
config.ini
that is not being used for a
MySQL server or other application for this reason (see
Section聽15.4.4.6, 鈥淒efining SQL and Other API Nodes鈥).
You can verify that ndb_restore is connected to the cluster by using the SHOW command in the ndb_mgm management client. You can also accomplish this from a system shell, as shown here:
shell> ndb_mgm -e "SHOW"
-n
is used to specify the node ID of the
data node on which the backups were taken.
The first time you run the ndb_restore
restoration program, you also need to restore the metadata.
In other words, you must re-create the database tables
鈥 this can be done by running it with the
-m
option. Note that the cluster should
have an empty database when starting to restore a backup.
(In other words, you should start ndbd
with --initial
prior to performing the
restore. You should also remove manually any Disk Data files
present in the data node's DataDir
.)
It is possible to restore data without restoring table metadata. Prior to MySQL 5.1.17, ndb_restore did not perform any checks of table schemas; if a table was altered between the time the backup was taken and when ndb_restore was run, ndb_restore would still attempt to restore the data to the altered table.
Beginning with MySQL 5.1.17, the default behavior is for
ndb_restore is to fail with an error if
table data do not match the table schema; this can be
overridden using the --skip-table-check
or
-s
option. If this option is used, then
ndb_restore attempts to fit data into the
existing table schema. The result of restoring a
backup to a table schema that does not match the original is
unspecified and is subject to change without
notice. (Bug#24363)
The -b
option is used to specify the ID or
sequence number of the backup, and is the same number shown
by the management client in the Backup
message displayed upon completion of a backup. (See
Section聽15.8.2, 鈥淯sing The Management Client to Create a Backup鈥.)
backup_id
completed
-e
adds (or restores) epoch information to
the cluster replication status table. This is useful for
starting replication on a MySQL Cluster replication slave.
When this option is used, the row in the
mysql.ndb_apply_status
having
0
in the id
column is
updated if it already exists; such a row is inserted if it
does not already exist. (See
Section聽15.10.9, 鈥淢ySQL Cluster Backups With Replication鈥.)
The path to the backup directory is required, and must
include the subdirectory corresponding to the ID backup of
the backup to be restored. For example, if the data node's
DataDir
is
/var/lib/mysql-cluster
, then the backup
directory is
/var/lib/mysql-cluster/BACKUP
, and the
backup files for the backup with the ID 3 can be found in
/var/lib/mysql-cluster/BACKUP/BACKUP-3
.
The path may be absolute or relative to the directory in
which the ndb_restore executable is
located, and may be optionally prefixed with
backup_path=
.
When restoring cluster backups, you must be sure to restore all data nodes from backups having the same backup ID. Using files from different backups will at best result in restoring the cluster to an inconsistent state, and may fail altogether.
It is possible to restore a backup to a database with a
different configuration than it was created from. For
example, suppose that a backup with backup ID
12
, created in a cluster with two
database nodes having the node IDs 2
and
3
, is to be restored to a cluster with
four nodes. Then ndb_restore must be run
twice 鈥 once for each database node in the cluster
where the backup was taken. However,
ndb_restore cannot always restore backups
made from a cluster running one version of MySQL to a
cluster running a different MySQL version. See
Section聽15.5.2, 鈥淐luster Upgrade and Downgrade Compatibility鈥,
for more information.
For more rapid restoration, the data may be restored in
parallel, provided that there is a sufficient number of
cluster connections available. That is, when restoring to
multiple nodes in parallel, you must have an
[API]
or [MYSQLD]
section in the cluster config.ini
file available for each concurrent
ndb_restore process. However, the data
files must always be applied before the logs.
Most of the options available for this program are shown in the following table:
Long Form | Short Form | Description | Default Value |
--backup-id | -b | Backup sequence ID | 0 |
--backup_path | None | Path to backup files | ./ |
--character-sets-dir | None | Specify the directory where character set information can be found | None |
--connect , --connectstring , or
--ndb-connectstring | -c or -C | Set the connectstring in
[nodeid=
format | localhost:1186 |
--core-file | None | Write a core file in the event of an error | TRUE |
--debug | -# | Output debug log | d:t:O, |
--dont_ignore_systab_0 | -f | Do not ignore system table during restore 鈥 EXPERIMENTAL; not for production use | FALSE |
--help or --usage | -? | Display help message with available options and current values, then exit | [N/A] |
--ndb-mgmd-host | None | Set the host and port in
format for the management server to connect to; this
is the same as --connect ,
--connectstring , or
--ndb-connectstring , but without a
way to specify the nodeid | None |
--ndb-nodegroup-map | -z | Specifies a nodegroup map 鈥 Syntax: list of
(source_nodegroup ,
destination_nodegroup ) | None |
--ndb-nodeid | None | Specify a node ID for the ndb_restore process | 0 |
--ndb-optimized-node-selection | None | Optimize selection of nodes for transactions | TRUE |
--ndb-shm | None | Use shared memory connections when available | FALSE |
--no-restore-disk-objects | -d | Do not restore Disk Data objects such as tablespaces and log file groups | FALSE (in other words, restore Disk Data objects
unless this option is used) |
--nodeid | -n | Use backup files from node with the specified ID | 0 |
--parallelism | -p | Set from 1 to 1024 parallel transactions to be used during the restoration process | 128 |
--print | None | Print metadata and log to stdout | FALSE |
--print_data | None | Print data to stdout | FALSE |
--print_log | None | Print log to stdout | FALSE |
--print_meta | None | Print metadata to stdout | FALSE |
--restore_data | -r | Restore data and logs | FALSE |
--restore_epoch | -e | Restore epoch data into the status table; the row in the
cluster.apply_status having the
id 0 is inserted or updated as
appropriate 鈥 this is convenient when starting
up replication on a MySQL Cluster replication slave | FALSE |
--restore_meta | -m | Restore table metadata | FALSE |
--skip-table-check | -s | Do not check table schemas (Added in MySQL 5.1.17) | FALSE |
--version | -V | Output version information and exit | [N/A] |
Beginning with MySQL 5.1.18, several additional options are
available for use with the --print_data
option in generating data dumps, either to
stdout
, or to a file. These are similar
to some of the options used with
mysqldump, and are shown in the following
table:
Long Form | Short Form | Description | Default Value |
--tab | -T | Creates dumpfiles, one per table, each named
.
Takes as its argument the path to the directory
where the files should be saved (required; use
. for the current directory). | None |
--fields-enclosed-by | None | String used to enclose all column values | None |
--fields-optionally-enclosed-by | None | String used to enclose column values containing character data (such as
CHAR ,
VARCHAR ,
BINARY ,
TEXT , or
ENUM ) | None |
--fields-terminated-by | None | String used to separate column values | \t (tab character) |
--hex | None | Use hex format for binary values | [N/A] |
--lines-terminated-by | None | String used to terminate each line | \n (linefeed character) |
--appends | None | When used with --tab , causes the data to be appended to
existing files of the same name | [N/A] |
If a table has no explicit primary key, then the output
generated when using the --print
includes
the table's hidden primary key.
Beginning with MySQL 5.1.18, it is possible to restore selected databases, or to restore selected tables from a given database using the syntax shown here:
ndb_restoreother_options
db_name_1
[db_name_2
[,db_name_3
][, ...] |tbl_name_1
[,tbl_name_2
][, ...]]
In other words, you can specify either of the following to be restored:
All tables from one or more databases
One or more tables from a single database
Note:
ndb_restore reports both temporary and
permanent errors. In the case of temporary errors, it may
able to recover from them. Beginning with MySQL 5.1.12, it
reports Restore successful, but encountered
temporary error, please look at configuration
in
such cases.
Five configuration parameters are essential for backup:
BackupDataBufferSize
The amount of memory used to buffer data before it is written to disk.
BackupLogBufferSize
The amount of memory used to buffer log records before these are written to disk.
BackupMemory
The total memory allocated in a database node for backups. This should be the sum of the memory allocated for the backup data buffer and the backup log buffer.
BackupWriteSize
The default size of blocks written to disk. This applies for both the backup data buffer and the backup log buffer.
BackupMaxWriteSize
The maximum size of blocks written to disk. This applies for both the backup data buffer and the backup log buffer.
More detailed information about these parameters can be found in Backup Parameters.
If an error code is returned when issuing a backup request, the
most likely cause is insufficient memory or disk space. You
should check that there is enough memory allocated for the
backup. Important: If you have
set BackupDataBufferSize
and
BackupLogBufferSize
and their sum is greater
than 4MB, then you must also set BackupMemory
as well. See
BackupMemory
.
You should also make sure that there is sufficient space on the hard drive partition of the backup target.
NDB
does not support repeatable reads, which
can cause problems with the restoration process. Although the
backup process is 鈥hot鈥, restoring a MySQL Cluster
from backup is not a 100% 鈥hot鈥 process. This is
due to the fact that, for the duration of the restore process,
running transactions get non-repeatable reads from the restored
data. This means that the state of the data is inconsistent
while the restore is in progress.
This section discusses the MySQL Cluster utility programs that can
be found in the mysql/bin
directory. Each of
these 鈥 except for ndb_size.pl
and
ndb_error_reporter 鈥 is a standalone
binary that can be used from a system shell, and that does not
need to connect to a MySQL server (nor even requires that a MySQL
server be connected to the cluster).
These utilities can also serve as examples for writing your own
applications using the NDB
API. The source code
for most of these programs may be found in the
storage/ndb/tools
directory of the MySQL
5.1 tree (see Section聽2.9, 鈥淢ySQL Installation Using a Source Distribution鈥).
The NDB
API is not covered in this manual;
please refer to the
NDB
API
Guide for information about this API.
All of the NDB
utilities are listed here with
brief descriptions:
ndb_config: Retrieves Cluster configuration option values.
ndb_cpcd: Used in testing and debugging MySQL Cluster.
ndb_delete_all: Deletes all rows from a given table.
ndb_desc: Lists all properties of an
NDB
table.
ndb_drop_index: Drops the specified index
from an NDB
table.
ndb_drop_table: Drops an
NDB
table.
ndb_error_reporter: Can be used to gather information useful for diagnosing problems with the cluster.
ndb_mgm: This is the MySQL Cluster management client, which is discussed in Section聽15.7.2, 鈥淐ommands in the MySQL Cluster Management Client鈥.
ndb_print_backup_file: Prints diagnostic information obtained from cluster backup files.
ndb_print_schema_file: Prints diagnostic information obtained from cluster schema files.
ndb_print_sys_file: Prints diagnostic information obtained from cluster system files.
ndb_restore: This utility is used to restore a cluster from backup. See Section聽15.8.3, 鈥ndb_restore 鈥 Restore a Cluster Backup鈥, for more information.
ndb_select_all: Prints all rows from an
NDB
table.
ndb_select_count: Gets the number of rows
in one or more NDB
tables.
ndb_show_tables: Shows all
NDB
tables anywhere in the cluster.
ndb_size.pl: Examines all the tables in a
given non-Cluster database and calculates the amount of
storage each would require if it were converted to use the
NDB
storage engine.
ndb_waiter: Reports on the status of
cluster data nodes in a manner similar to that of the
management client command ALL STATUS
.
Most of these utilities need to connect to a Cluster management
server in order to function. The exceptions are
ndb_size.pl
(see below), and the following
utilities which access a cluster data node filesystem and so need
to be run on a data node host:
ndb_print_backup_file
ndb_print_schema_file
ndb_print_sys_file
ndb_size.pl
is a Perl script which is also
intended to be used from the shell; however it is a MySQL
application and must be able to connect to a MySQL server. See
Section聽15.9.14, 鈥ndb_size.pl 鈥 NDBCluster Size Requirement Estimator鈥, for additional
requirements for using this script.
ndb_error_reporter is also a Perl script. It is used to gather cluster data node and management node logs together into a tarball to submit along with a bug report. It can use ssh or scp to access the node filesystems remotely.
Additional information about each of these utilities (except for ndb_mgm and ndb_restore) can be found in the sections that follow.
Note: All of these utilities (except for ndb_size.pl and ndb_config) can use the options discussed in Section聽15.6.5, 鈥淐ommand Options for MySQL Cluster Processes鈥. Additional options specific to each utility program are discussed in the individual program listings.
The order in which these options are used is generally not important. For example, all of these commands produce exactly the same output:
ndb_desc -c localhost fish -d test
ndb_desc fish -c localhost -d test
ndb_desc -d test fish -c localhost
This tool extracts configuration information for data nodes,
SQL nodes, and API nodes from a cluster management node (and
possibly its config.ini
file).
Usage:
ndb_config options
The options
available for this
utility differ somewhat from those used with the other
utilities, and so are listed in their entirety in the next
section, followed by some examples.
Options:
Causes ndb_config to print a list of available options, and then exit.
Causes ndb_config to print a version information string, and then exit.
--ndb-connectstring=
connect_string
Specifies the connectstring to use in connecting to the
management server. The format for the connectstring is
the same as described in
Section聽15.4.4.2, 鈥淭he Cluster Connectstring鈥, and
defaults to localhost:1186
.
The use of -c
as a short version for
this option is supported for
ndb_config beginning with MySQL
5.1.12.
Gives the path to the management server's configuration
file (config.ini
). This may be a
relative or absolute path. If the management node
resides on a different host from the one on which
ndb_config is invoked, then an
absolute path must be used.
--query=
,
query-options
-q
query-options
This is a comma-delimited list of query
options 鈥 that is, a list of one or
more node attributes to be returned. These include
id
(node ID), type (node type 鈥
that is, ndbd
,
mysqld
, or
ndb_mgmd
), and any configuration
parameters whose values are to be obtained.
For example,
--query=id,type,indexmemory,datamemory
would return the node ID, node type,
DataMemory
, and
IndexMemory
for each node.
Note: If a given parameter is not applicable to a certain type of node, than an empty string is returned for the corresponding value. See the examples later in this section for more information.
Specifies the hostname of the node for which configuration information is to be obtained.
--id=
,
node_id
--nodeid=
node_id
Used to specify the node ID of the node for which configuration information is to be obtained.
(Tells ndb_config to print
information from parameters defined in
[ndbd]
sections only. Currently,
using this option has no affect, since these are the
only values checked, but it may become possible in
future to query parameters set in
[tcp]
and other sections of cluster
configuration files.)
Filters results so that only configuration values
applying to nodes of the specified
node_type
(ndbd
, mysqld
, or
ndb_mgmd
) are returned.
--fields=
,
delimiter
-f
delimiter
Specifies a delimiter
string
used to separate the fields in the result. The default
is 鈥,
鈥 (the comma
character).
Note: If the
delimiter
contains spaces or
escapes (such as \n
for the linefeed
character), then it must be quoted.
--rows=
,
separator
-r
separator
Specifies a separator
string
used to separate the rows in the result. The default is
a space character.
Note: If the
separator
contains spaces or
escapes (such as \n
for the linefeed
character), then it must be quoted.
Examples:
To obtain the node ID and type of each node in the cluster:
shell> ./ndb_config --query=id,type --fields=':' --rows='\n'
1:ndbd
2:ndbd
3:ndbd
4:ndbd
5:ndb_mgmd
6:mysqld
7:mysqld
8:mysqld
9:mysqld
In this example, we used the --fields
options to separate the ID and type of each node with a
colon character (:
), and the
--rows
options to place the values for
each node on a new line in the output.
To produce a connectstring that can be used by data, SQL, and API nodes to connect to the management server:
shell> ./ndb_config --config-file=usr/local/mysql/cluster-data/config.ini --query=hostname,portnumber --fields=: --rows=, --type=ndb_mgmd
192.168.0.179:1186
This invocation of ndb_config checks
only data nodes (using the --type
option), and shows the values for each node's ID and
hostname, and its DataMemory
,
IndexMemory
, and
DataDir
parameters:
shell> ./ndb_config --type=ndbd --query=id,host,datamemory,indexmemory,datadir -f ' : ' -r '\n'
1 : 192.168.0.193 : 83886080 : 18874368 : /usr/local/mysql/cluster-data
2 : 192.168.0.112 : 83886080 : 18874368 : /usr/local/mysql/cluster-data
3 : 192.168.0.176 : 83886080 : 18874368 : /usr/local/mysql/cluster-data
4 : 192.168.0.119 : 83886080 : 18874368 : /usr/local/mysql/cluster-data
In this example, we used the short options
-f
and -r
for setting
the field delimiter and row separator, respectively.
To exclude results from any host except one in
particular, use the --host
option:
shell> ./ndb_config --host=192.168.0.176 -f : -r '\n' -q id,type
3:ndbd
5:ndb_mgmd
In this example, we also used the short form
-q
to determine the attributes to be
queried.
Similarly, you can limit results to a node with a
specific ID using the --id
or
--nodeid
option.
This utility is found in the libexec
directory. It is part of an internal automated test
framework used in testing and bedugging MySQL Cluster.
Because it can control processes on remote systems, it is
not advisable to use ndb_cpcd in a
production cluster.
Because some users may be interested in employing the Cluster testing framework for their own development or testing purposes, we intend to make details of this application's usage available in the near future as part of the MySQL Internals Manual.
The source files for ndb_cpcd may be
found in the directory
storage/ndb/src/cw/cpcd
, in the MySQL
5.1 source tree.
ndb_delete_all deletes all rows from the
given NDB
table. In some cases, this can
be much faster than DELETE
or even
TRUNCATE
.
Usage:
ndb_delete_all -cconnect_string
tbl_name
-ddb_name
This deletes all rows from the table named
tbl_name
in the database named
db_name
. It is exactly equivalent
to executing TRUNCATE
in MySQL.
db_name
.tbl_name
Additional Options:
ndb_desc provides a detailed description
of one or more NDB
tables.
Usage:
ndb_desc -cconnect_string
tbl_name
-ddb_name
Sample Output:
MySQL table creation and population statements:
USE test; CREATE TABLE fish ( id INT(11) NOT NULL AUTO_INCREMENT, name VARCHAR(20), PRIMARY KEY pk (id), UNIQUE KEY uk (name) ) ENGINE=NDBCLUSTER; INSERT INTO fish VALUES ('','guppy'), ('','tuna'), ('','shark'), ('','manta ray'), ('','grouper'), ('','puffer');
Output from ndb_desc:
shell> ./ndb_desc -c localhost fish -d test -p
-- fish --
Version: 16777221
Fragment type: 5
K Value: 6
Min load factor: 78
Max load factor: 80
Temporary table: no
Number of attributes: 2
Number of primary keys: 1
Length of frm data: 268
Row Checksum: 1
Row GCI: 1
TableStatus: Retrieved
-- Attributes --
id Int PRIMARY KEY DISTRIBUTION KEY AT=FIXED ST=MEMORY
name Varchar(20;latin1_swedish_ci) NULL AT=SHORT_VAR ST=MEMORY
-- Indexes --
PRIMARY KEY(id) - UniqueHashIndex
uk(name) - OrderedIndex
PRIMARY(id) - OrderedIndex
uk$unique(name) - UniqueHashIndex
-- Per partition info --
Partition Row count Commit count Frag fixed memory Frag varsized memory
2 2 2 65536 327680
1 2 2 65536 327680
3 2 2 65536 327680
NDBT_ProgramExit: 0 - OK
Additional Options:
ndb_drop_index drops the specified index
from an NDB
table. It is
recommended that you use this utility only as an example for
writing NDB API applications 鈥 see the
Warning later in this section for details.
Usage:
ndb_drop_index -cconnect_string
table_name
index
-ddb_name
The statement shown above drops the index named
index
from the
table
in the
database
.
Additional Options: None that are specific to this application.
Warning: Operations performed on Cluster table indexes using the NDB API are not visible to MySQL and make the table unusable by a MySQL server. If you use this program to drop an index, then try to access the table from an SQL node, an error results, as shown here:
shell>./ndb_drop_index -c localhost dogs ix -d ctest1
Dropping index dogs/idx...OK NDBT_ProgramExit: 0 - OK shell>./mysql -u jon -p ctest1
Enter password: ******* Reading table information for completion of table and column names You can turn off this feature to get a quicker startup with -A Welcome to the MySQL monitor. Commands end with ; or \g. Your MySQL connection id is 7 to server version: 5.1.12-beta-20060817 Type 'help;' or '\h' for help. Type '\c' to clear the buffer. mysql> SHOW TABLES; +------------------+ | Tables_in_ctest1 | +------------------+ | a | | bt1 | | bt2 | | dogs | | employees | | fish | +------------------+ 6 rows in set (0.00 sec) mysql> SELECT * FROM dogs; ERROR 1296 (HY000): Got error 4243 'Index not found' from NDBCLUSTER
In such a case, your only option for
making the table available to MySQL again is to drop the
table and re-create it. You can use either the SQL
statementDROP TABLE
or the
ndb_drop_table utility (see
Section聽15.9.6, 鈥ndb_drop_table 鈥 Drop NDB Table鈥) to
drop the table.
ndb_drop_table drops the specified
NDB
table. (If you try to use this on a
table created with a storage engine other than NDB, it fails
with the error 723: No such table
exists.) This operation is extremely fast
鈥 in some cases, it can be an order of magnitude
faster than using DROP TABLE
on an
NDB
table from MySQL.
Usage:
ndb_drop_table -cconnect_string
tbl_name
-ddb_name
Additional Options: None.
ndb_error_reporter creates an archive from data node and management node log files that can be used to help diagnose bugs or other problems with a cluster. It is highly recommended that you make use of this utility when filing reports of bugs in MySQL Cluster.
Usage:
ndb_error_reporterpath/to/config-file
[username
] [--fs]
This utility is intended for use on a management node host,
and requires the path to the management host configuration
file (config.ini
). Optionally, you can
supply the name of a user that is able to access the
cluster's data nodes via SSH, in order to copy the data node
log files. ndb_error_reporter then includes all of these
files in archive that is created in the same directory in
which it is run. The archive is named
ndb_error_report_
,
where YYYYMMDDHHMMSS
.tar.bz2YYYYMMDDHHMMSS
is a
datetime string.
If the --fs
is used, then the data node
filesystems are also copied to the management host and
included in the archive that is produced by this script. As
data node filesystems can be extremely large even after
being compressed, we ask that you please do
not send archives created using this
option to MySQL AB unless you are specifically requested to
do so.
ndb_print_backup_file obtains diagnostic information from a cluster backup file.
Usage:
ndb_print_backup_file file_name
file_name
is the name of a
cluster backup file. This can be any of the files
(.Data
, .ctl
, or
.log
file) found in a cluster backup
directory. These files are found in the data node's backup
directory under the subdirectory
BACKUP-
,
where #
#
is the sequence number
for the backup. For more information about cluster backup
files and their contents, see
Section聽15.8.1, 鈥淐luster Backup Concepts鈥.
Like ndb_print_schema_file and
ndb_print_sys_file (and unlike most of
the other NDB
utilities that are intended
to be run on a management server host or to connect to a
management server) ndb_print_backup_file
must be run on a cluster data node, since it accesses the
data node filesystem directly. Because it does not make use
of the management server, this utility can be used when the
management server is not running, and even when the cluster
has been completely shut down.
Additional Options: None.
ndb_print_schema_file obtains diagnostic information from a cluster schema file.
Usage:
ndb_print_schema_file file_name
file_name
is the name of a
cluster schema file.
Like ndb_print_backup_file and
ndb_print_sys_file (and unlike most of
the other NDB
utilities that are intended
to be run on a management server host or to connect to a
management server) ndb_schema_backup_file
must be run on a cluster data node, since it accesses the
data node filesystem directly. Because it does not make use
of the management server, this utility can be used when the
management server is not running, and even when the cluster
has been completely shut down.
Additional Options: None.
ndb_print_sys_file obtains diagnostic information from a cluster system file.
Usage:
ndb_print_sys_file file_name
file_name
is the name of a
cluster system file (sysfile). Cluster system files are
located in a data node's data directory
(DataDir
); the path under this directory
to system files matches the pattern
ndb_
.
In each case, the #
_fs/D#
/DBDIH/P#
.sysfile#
represents a
number (not necessarily the same number).
Like ndb_print_backup_file and
ndb_print_schema_file (and unlike most of
the other NDB
utilities that are intended
to be run on a management server host or to connect to a
management server) ndb_print_backup_file
must be run on a cluster data node, since it accesses the
data node filesystem directly. Because it does not make use
of the management server, this utility can be used when the
management server is not running, and even when the cluster
has been completely shut down.
Additional Options: None.
ndb_select_all prints all rows from an
NDB
table to stdout
.
Usage:
ndb_select_all -cconnect_string
tbl_name
-ddb_name
[>file_name
]
Additional Options:
--lock=
,
lock_type
-l
lock_type
Employs a lock when reading the table. Possible values
for lock_type
are:
0
: Read lock
1
: Read lock with hold
2
: Exclusive read lock
There is no default value for this option.
--order=
,
index_name
-o
index_name
Orders the output according to the index named
index_name
. Note that this is
the name of an index, not of a column, and that the
index must have been explicitly named when created.
Sorts the output in descending order. This option can be
used only in conjunction with the -o
(--order
) option.
Excludes column headers from the output.
Causes all numeric values to be displayed in hexadecimal format. This does not affect the output of numerals contained in strings or datetime values.
--delimiter=
,
character
-D
character
Causes the character
to be
used as a column delimiter. Only table data columns are
separated by this delimiter.
The default delimiter is the tab character.
Adds a disk reference column to the output. The column is non-empty only for Disk Data tables having non-indexed columns.
Adds a ROWID
column providing
information about the fragments in which rows are
stored.
Adds a column to the output showing the global checkpoint at which each row was last updated. See Section聽15.15, 鈥淢ySQL Cluster Glossary鈥, and Section聽15.7.3.2, 鈥淟og Events鈥, for more information about checkpoints.
Scan the table in the order of the tuples.
Causes any table data to be omitted.
Sample Output:
Output from a MySQL SELECT
statement:
mysql> SELECT * FROM ctest1.fish;
+----+-----------+
| id | name |
+----+-----------+
| 3 | shark |
| 6 | puffer |
| 2 | tuna |
| 4 | manta ray |
| 5 | grouper |
| 1 | guppy |
+----+-----------+
6 rows in set (0.04 sec)
Output from the equivalent invocation of ndb_select_all:
shell> ./ndb_select_all -c localhost fish -d ctest1
id name
3 [shark]
6 [puffer]
2 [tuna]
4 [manta ray]
5 [grouper]
1 [guppy]
6 rows returned
NDBT_ProgramExit: 0 - OK
Note that all string values are enclosed by square brackets
(鈥[
...]
鈥)
in the output of ndb_select_all. For a
further example, consider the table created and populated as
shown here:
CREATE TABLE dogs ( id INT(11) NOT NULL AUTO_INCREMENT, name VARCHAR(25) NOT NULL, breed VARCHAR(50) NOT NULL, PRIMARY KEY pk (id), KEY ix (name) ) TABLESPACE ts STORAGE DISK ENGINE=NDB; INSERT INTO dogs VALUES ('', 'Lassie', 'collie'), ('', 'Scooby-Doo', 'Great Dane'), ('', 'Rin-Tin-Tin', 'Alsatian'), ('', 'Rosscoe', 'Mutt');
This demonstrates the use of several additional ndb_select_all options:
shell> ./ndb_select_all -d ctest1 dogs -o ix -z --gci --disk
GCI id name breed DISK_REF
834461 2 [Scooby-Doo] [Great Dane] [ m_file_no: 0 m_page: 98 m_page_idx: 0 ]
834878 4 [Rosscoe] [Mutt] [ m_file_no: 0 m_page: 98 m_page_idx: 16 ]
834463 3 [Rin-Tin-Tin] [Alsatian] [ m_file_no: 0 m_page: 34 m_page_idx: 0 ]
835657 1 [Lassie] [Collie] [ m_file_no: 0 m_page: 66 m_page_idx: 0 ]
4 rows returned
NDBT_ProgramExit: 0 - OK
ndb_select_count prints the number of
rows in one or more NDB
tables. With a
single table, the result is equivalent to that obtained by
using the MySQL statement SELECT COUNT(*) FROM
.
tbl_name
Usage:
ndb_select_count [-cconnect_string
] -ddb_name
tbl_name
[,tbl_name2
[, ...]]
Additional Options: None that are specific to this application. However, you can obtain row counts from multiple tables in the same database by listing the table names separated by spaces when invoking this command, as shown under Sample Output.
Sample Output:
shell> ./ndb_select_count -c localhost -d ctest1 fish dogs
6 records in table fish
4 records in table dogs
NDBT_ProgramExit: 0 - OK
ndb_show_tables displays a list of all
NDB
database objects in the cluster. By
default, this includes not only both user-created tables and
NDB
system tables, but
NDB
-specific indexes, internal triggers,
and Cluster Disk Data objects as well.
Usage:
ndb_show_tables [-c connect_string
]
Additional Options:
Specifies the number of times the utility should execute. This is 1 when this option is not specified, but if you do use the option, you must supply an integer argument for it.
Using this option causes the output to be in a format
suitable for use with LOAD DATA
INFILE
.
Can be used to restrict the output to one type of object, specified by an integer type code as shown here:
1: System table
2: User-created table
3: Unique hash index
Any other value causes all NDB
database objects to be listed (the default).
If specified, this causes unqualified object names to be displayed.
Note: Only user-created
Cluster tables may be accessed from MySQL; system tables
such as SYSTAB_0
are not visible to
mysqld. However, you can examine the
contents of system tables using NDB
API
applications such as ndb_select_all (see
Section聽15.9.11, 鈥ndb_select_all 鈥 Print Rows from NDB Table鈥).
This is a Perl script that can be used to estimate the
amount of space that would be required by a MySQL database
if it were converted to use the
NDBCluster
storage engine. Unlike the
other utilities discussed in this section, it does not
require access to a MySQL Cluster (in fact, there is no
reason for it to do so). However, it does need to access the
MySQL server on which the database to be tested resides.
Requirements:
A running MySQL server. The server instance does not have to provide support for MySQL Cluster.
A working installation of Perl.
The DBI
and
HTML::Template
modules, both of which
can be obtained from CPAN if they are not already part
of your Perl installation. (Many Linux and other
operating system distribution provide their own packages
for one or both of these libraries.)
The ndb_size.tmpl
template file,
which you should be able to find in the
share/mysql
directory of your MySQL
installation. This file should be copied or moved into
the same directory as ndb_size.pl
鈥 if it is not there already 鈥 before
running the script.
A MySQL user account having the necessary privileges. If
you do not wish to use an existing account, then
creating one using GRANT USAGE ON
鈥
where db_name
.*db_name
is the name of
the database to be examined 鈥 is sufficient for
this purpose.
ndb_size.pl
and
ndb_size.tmpl
can also be found in the
MySQL sources in storage/ndb/tools
. If
these files are not present in your MySQL installation, you
can obtain them from the
MySQLForge
project page.
Usage:
perl ndb_size.pldb_name
hostname
username
password
>file_name
.html
The command shown connects to the MySQL server at
hostname
using the account of the
user username
having the password
password
, analyses all of the
tables in database db_name
, and
generates a report in HTML format which is directed to the
file
.
(Without the redirection, the output is sent to
file_name
.htmlstdout
.) This figure shows partial sample
output as viewed in a Web browser:
The output from this script includes:
Minimum values for the DataMemory
,
IndexMemory
,
MaxNoOfTables
,
MaxNoOfAttributes
,
MaxNoOfOrderedIndexes
,
MaxNoOfUniqueHashIndexes
, and
MaxNoOfTriggers
configuration
parameters required to accommodate the tables analysed.
Memory requirements for all of the tables, attributes, ordered indexes, and unique hash indexes defined in the database.
The IndexMemory
and
DataMemory
required per table and
table row.
ndb_waiter repeatedly (each 100
milliseconds) prints out the status of all cluster data
nodes until either the cluster reaches a given status or the
--timeout
limit is exceeded, then exits. By
default, it waits for the cluster to achieve
STARTED
status, in which all nodes have
started and connected to the cluster. This can be overridden
using the --no-contact
and
--not-started
options (see
Additional
Options).
The node states reported by this utility are as follows:
NO_CONTACT
: The node cannot be
contacted.
UNKNOWN
: The node can be contacted,
but its status is not yet known. Usually, this means
that the node has received a START
or
RESTART
command from the management
server, but has not yet acted on it.
NOT_STARTED
: The node has stopped,
but remains in contact with the cluster. This is seen
when restarting the node using the management client's
RESTART
command.
STARTING
: The node's
ndbd process has started, but the
node has not yet joined the cluster.
STARTED
: The node is operational, and
has joined the cluster.
SHUTTING_DOWN
: The node is shutting
down.
SINGLE USER MODE
: This is shown for
all cluster data nodes when the cluster is in single
user mode.
Usage:
ndb_waiter [-c connect_string
]
Instead of waiting for the STARTED
state, ndb_waiter continues running
until the cluster reaches NO_CONTACT
status before exiting.
Instead of waiting for the STARTED
state, ndb_waiter continues running
until the cluster reaches NOT_STARTED
status before exiting.
Time to wait. The program exits if the desired state is not achieved within this number of seconds. The default is 120 seconds (1200 reporting cycles).
Sample Output:
Shown here is the output from ndb_waiter
when run against a 4-node cluster in which two nodes have
been shut down and then started again manually. Duplicate
reports (indicated by 鈥...
鈥)
are omitted.
shell> ./ndb_waiter -c localhost
Connecting to mgmsrv at (localhost)
State node 1 STARTED
State node 2 NO_CONTACT
State node 3 STARTED
State node 4 NO_CONTACT
Waiting for cluster enter state STARTED
...
State node 1 STARTED
State node 2 UNKNOWN
State node 3 STARTED
State node 4 NO_CONTACT
Waiting for cluster enter state STARTED
...
State node 1 STARTED
State node 2 STARTING
State node 3 STARTED
State node 4 NO_CONTACT
Waiting for cluster enter state STARTED
...
State node 1 STARTED
State node 2 STARTING
State node 3 STARTED
State node 4 UNKNOWN
Waiting for cluster enter state STARTED
...
State node 1 STARTED
State node 2 STARTING
State node 3 STARTED
State node 4 STARTING
Waiting for cluster enter state STARTED
...
State node 1 STARTED
State node 2 STARTED
State node 3 STARTED
State node 4 STARTING
Waiting for cluster enter state STARTED
...
State node 1 STARTED
State node 2 STARTED
State node 3 STARTED
State node 4 STARTED
Waiting for cluster enter state STARTED
NDBT_ProgramExit: 0 - OK
Note: If no connectstring
is specified, then ndb_waiter tries to
connect to a management on localhost
, and
reports Connecting to mgmsrv at (null)
.
Previous to MySQL 5.1.6, asynchronous replication, more usually referred to simply as 鈥replication鈥, was not available when using MySQL Cluster. MySQL 5.1.6 introduces master-slave replication of this type for MySQL Cluster databases. This section explains how to set up and manage a configuration wherein one group of computers operating as a MySQL Cluster replicates to a second computer or group of computers. We assume some familiarity on the part of the reader with standard MySQL replication as discussed elsewhere in this Manual. (See Chapter聽6, Replication).
Normal (non-clustered) replication involves a
鈥master鈥 server and a 鈥slave鈥 server,
the master being the source of the operations and data to be
replicated and the slave being the recipient of these. In MySQL
Cluster, replication is conceptually very similar but can be more
complex in practice, as it may be extended to cover a number of
different configurations including replicating between two
complete clusters. Although a MySQL Cluster itself depends on the
NDB Cluster
storage engine for clustering
functionality, it is not necessary to use the Cluster storage
engine on the slave. However, for maximum availability, it is
possible to replicate from one MySQL Cluster to another, and it is
this type of configuration that we discuss, as shown in the
following figure:
In this scenario, the replication process is one in which
successive states of a master cluster are logged and saved to a
slave cluster. This process is accomplished by a special thread
known as the NDB binlog injector thread, which runs on each MySQL
server and produces a binary log (binlog
). This
thread ensures that all changes in the cluster producing the
binary log 鈥 and not just those changes that are effected
via the MySQL Server 鈥 are inserted into the binary log with
the correct serialization order. We refer to the MySQL replication
master and replication slave servers as replication servers or
replication nodes, and the data flow or line of communication
between them as a replication channel.
Throughout this section, we use the following abbreviations or symbols for referring to the master and slave clusters, and to processes and commands run on the clusters or cluster nodes:
Symbol or Abbreviation | Description (Refers to...) |
M | The cluster serving as the (primary) replication master |
S | The cluster acting as the (primary) replication slave |
shell | Shell command to be issued on the master cluster |
mysql | MySQL client command issued on a single MySQL server running as an SQL node on the master cluster |
mysql | MySQL client command to be issued on all SQL nodes participating in the replication master cluster |
shell | Shell command to be issued on the slave cluster |
mysql | MySQL client command issued on a single MySQL server running as an SQL node on the slave cluster |
mysql | MySQL client command to be issued on all SQL nodes participating in the replication slave cluster |
C | Primary replication channel |
C' | Secondary replication channel |
M' | Secondary replication master |
S' | Secondary replication slave |
A replication channel requires two MySQL servers acting as replication servers (one each for the master and slave). For example, this means that in the case of a replication setup with two replication channels (to provide an extra channel for redundancy), there will be a total of four replication nodes, two per cluster.
Replication of a MySQL Cluster as described in this section and
those following is dependent on row-based replication. This
means that the replication master MySQL server must be started
with --binlog-format=row
, as described in
Section聽15.10.6, 鈥淪tarting Replication (Single Replication Channel)鈥. For
general information about row-based replication, see
Section聽6.1.2, 鈥淩eplication Formats鈥.
(It is possible to replicate a MySQL Cluster using statement-based replication. However, in this case, the following restrictions apply: All updates to data rows on the cluster acting as the master must be directed to a single MySQL server; It is not possible to replicate a cluster using several MySQL replication processes at the same time; Only changes made at the SQL level are replicated.)
Each MySQL server used for replication in either cluster must be
uniquely identified among all the MySQL replication servers
participating in either cluster (you cannot have replication
servers on both the master and slave clusters sharing the same
ID). This can be done by starting each SQL node using the
--server-id=
option, where id
id
is a unique integer.
Although it is not strictly necessary, we will assume for
purposes of this discussion that all MySQL installations are the
same version.
In any event, both MySQL servers involved in replication must be compatible with one another with respect to both the version of the replication protocol used and the SQL feature sets which they support; the simplest and easiest way to assure that this is the case is to use the same MySQL version for all servers involved. Note that in many cases it is not possible to replicate to a slave running a version of MySQL with a lower version number than that of the master 鈥 see Section聽6.3.2, 鈥淩eplication Compatibility Between MySQL Versions鈥, for details.
We assume that the slave server or cluster is dedicated to replication of the master, and that no other data is being stored on it.
The following are known problems or issues when using replication with MySQL Cluster in MySQL 5.1:
Prior to MySQL 5.1.18, a MySQL Cluster replication slave mysqld had no way of detecting that the connection from the master had been interrupted (due to, for instance, the master going down or a network failure). For this reason, it was possible for the slave to become inconsistent with the master.
(If high availability is a requirement for the slave server or cluster, then it is still advisable to set up multiple replication lines, to monitor the master mysqld on the primary replication line, and to fail over to a secondary line if and as necessary. For information about implementing this type of setup, see Section聽15.10.7, 鈥淯sing Two Replication Channels鈥, and Section聽15.10.8, 鈥淚mplementing Failover with MySQL Cluster鈥.)
Beginning with MySQL 5.1.18, the master issues a
鈥gap鈥 event when connecting to the cluster.
When the slave encounters a gap in the replication log, it
stops with an error message. This message is available in
the output of SHOW SLAVE STATUS
, and
indicates that the SQL thread has stopped due to an incident
registered in the replication stream, and that manual
intervention is required. In order to restart the slave, it
is necessary to issue the following commands:
SET GLOBAL SQL_SLAVE_SKIP_COUNTER = 1; START SLAVE;
The slave then resumes reading the master binlog from the point where the gap was recorded.
Circular replication:
Prior to MySQL 5.1.18, circular replication was not supported with MySQL Cluster replication, due to the fact that all log events created in a particular MySQL Cluster were wrongly tagged with the server ID of the MySQL server used as master and not with the server ID of the originating server.
Beginning with MySQL 5.1.18, this limitation is lifted, as discussed in the next few paragraphs, in which we consider the example of a replication setup involving three MySQL Clusters numbered 1, 2, and 3, in which Cluster 1 acts as the replication master for Cluster 2, Cluster 2 acts as the master for Cluster 3, and Cluster 3 acts as the master for Cluster 1. Each cluster has two SQL nodes, with SQL nodes A and B belonging to Cluster 1, SQL nodes C and D belonging to Cluster 2, and SQL nodes E and F belonging to Cluster 3.
Circular replication using these clusters is supported as long as:
the SQL nodes on all masters and slaves are the same
All SQL nodes acting as replication masters and slaves
are started using the
--log-slave-updates
option
This type of circular replication setup is shown in the following diagram:
In this scenario, SQL node A in Cluster 1 replicates to SQL node C in Cluster 2; SQL node C replicates to SQL node E in Cluster 3; SQL node E replicates to SQL node A. In other words, the replication line (indicated by the red arrows in the diagram) directly connects all SQL nodes used as replication masters and slaves.
It should also be possible to set up circular replication in which not all master SQL nodes are also slaves, as shown here:
In this case, different SQL nodes in each cluster are used
as replication masters and slaves. However, you must
not start any of the SQL nodes using
--log-slave-updates
(see the
description of this
option for more information). This type of circular
replication scheme for MySQL Cluster, in which the line of
replication (again indicated by the red arrows in the
diagram) is discontinuous, should be possible, but it should
be noted that it has not yet been thoroughly tested and must
therefore still be considered experimental.
The use of data definition statements, such as
CREATE TABLE
, DROP
TABLE
, and ALTER TABLE
, are
recorded in the binary log for only the MySQL server on
which they are issued.
In MySQL 5.1.6, only those NDB
tables
having explicit primary keys could be replicated. This
limitation was lifted in MySQL 5.1.7.
Restarting the cluster with the
--initial
option will cause the
sequence of GCI and epoch numbers to start over from
0
. (This is generally true of MySQL
Cluster and not limited to replication scenarios involving
Cluster.) The MySQL servers involved in replication should
in this case be restarted. After this, you should use the
RESET MASTER
and RESET
SLAVE
statements to clear the invalid
ndb_binlog_index
and
ndb_apply_status
tables. respectively.
Trying to set values for the
auto_increment_offset
and
auto_increment_increment
server system
variables produces unpredictable results. The use of these
variables is not supported with MySQL Cluster replication.
Replication in MySQL Cluster makes use of a number of dedicated
tables in the mysql
database on each MySQL
Server instance acting as an SQL node in both the cluster being
replicated and the replication slave (whether the slave is a
single server or a cluster). These tables are created during the
MySQL installation process by the
mysql_install_db script, and include a table
for storing the binary log's indexing data. Since the
ndb_binlog_index
table is local to each MySQL
server and does not participate in clustering, it uses the
MyISAM
storage engine. This means that it
must be created separately on each mysqld
participating in the master cluster. (However, the binlog itself
contains updates from all MySQL servers in the cluster to be
replicated.) This table is defined as follows:
CREATE TABLE `ndb_binlog_index` ( `Position` BIGINT(20) UNSIGNED NOT NULL, `File` VARCHAR(255) NOT NULL, `epoch` BIGINT(20) UNSIGNED NOT NULL, `inserts` BIGINT(20) UNSIGNED NOT NULL, `updates` BIGINT(20) UNSIGNED NOT NULL, `deletes` BIGINT(20) UNSIGNED NOT NULL, `schemaops` BIGNINT(20) UNSIGNED NOT NULL, PRIMARY KEY (`epoch`) ) ENGINE=MYISAM DEFAULT CHARSET=latin1;
Important: Prior to MySQL
5.1.14, the ndb_binlog_index
table was known
as binlog_index
, and was kept in a separate
cluster
database, which in MySQL 5.1.7 and
earlier was known as the cluster_replication
database. Similarly, the ndb_apply_status
and
ndb_schema
tables were known as
apply_status
and schema
,
and were also found in the cluster
(earlier
cluster_replication
) database. However,
beginning with MySQL 5.1.14, all MySQL Cluster replication
tables reside in the mysql
system database.
Information about how this change affects upgrades from MySQL Cluster 5.1.13 and earlier to 5.1.14 and later versions can be found in Section聽C.1.5, 鈥淐hanges in release 5.1.14 (05 December 2006)鈥.
The following figure shows the relationship of the MySQL Cluster
replication master server, its binlog injector thread, and the
mysql.ndb_binlog_index
table.
An additional table, named ndb_apply_status
,
is used to keep a record of the operations that have been
replicated from the master to the slave. Unlike the case with
ndb_binlog_index
, the data in this table is
not specific to any one SQL node in the (slave) cluster, and so
ndb_apply_status
can use the NDB
Cluster
storage engine, as shown here:
CREATE TABLE `ndb_apply_status` ( `server_id` INT(10) UNSIGNED NOT NULL, `epoch` BIGINT(20) UNSIGNED NOT NULL, `log_name` VARCHAR(255) CHARACTER SET latin1 COLLATE latin1_bin NOT NULL, `start_pos` BIGINT(20) UNSIGNED NOT NULL, `end_pos` BIGINT(20) UNSIGNED NOT NULL, PRIMARY KEY (`server_id`) USING HASH ) ENGINE=NDBCLUSTER DEFAULT CHARSET=latin1;
The log_name
, start_pos
,
and end_pos
columns were added in MySQL
5.1.18.
If you are using MySQL Cluster replication, see Section聽15.5.2, 鈥淐luster Upgrade and Downgrade Compatibility鈥 before upgrading to MySQL 5.1.18 or later from an earlier version.
The ndb_binlog_index
and
ndb_apply_status
tables are created in the
mysql
database because they should not be
replicated. No user intervention is normally required to create
or maintain either of them. Both the
ndb_binlog_index
and the
ndb_apply_status
tables are maintained by the
NDB
injector thread. This keeps the master
mysqld process updated to changes performed
by the NDB
storage engine. The
NDB
binlog injector
thread receives events directly from the
NDB
storage engine. The
NDB
injector is responsible for capturing all
the data events within the cluster, and ensures that all events
which change, insert, or delete data are recorded in the
ndb_binlog_index
table. The slave I/O thread
transfers the events from the master's binary log to the slave's
relay log.
However, it is advisable to check for the existence and
integrity of these tables as an initial step in preparing a
MySQL Cluster for replication. It is possible to view event data
recorded in the binary log by querying the
mysql.binlog_index
table directly on the
master. This can be also be accomplished using the SHOW
BINLOG EVENTS
statement on either the replication
master or slave MySQL servers. (See
Section聽13.6.1.4, 鈥SHOW BINLOG EVENTS
Syntax鈥.)
You can also obtain useful information from the output of
SHOW ENGINE NDB
STATUS
.
Note: Beginning with MySQL
5.1.14, if the apply_status
table does not
exist on the slave, it is created by
ndb_restore. (Bug#14612)
Preparing the MySQL Cluster for replication consists of the following steps:
Check all MySQL servers for version compatibility (see Section聽15.10.2, 鈥淎ssumptions and General Requirements鈥).
Create a slave account on the master Cluster with the appropriate privileges:
mysqlM
>GRANT REPLICATION SLAVE
->ON *.* TO '
->slave_user
'@'slave_host
'IDENTIFIED BY '
slave_password
';
where slave_user
is the slave
account username, slave_host
is
the hostname or IP address of the replication slave, and
slave_password
is the password to
assign to this account.
For example, to create a slave user account with the name
鈥myslave
,鈥 logging in from
the host named 鈥rep-slave
,鈥
and using the password
鈥53cr37
,鈥 use the following
GRANT
statement:
mysqlM
>GRANT REPLICATION SLAVE
->ON *.* TO 'myslave'@'rep-slave'
->IDENTIFIED BY '53cr37';
For security reasons, it is preferable to use a unique user account 鈥 not employed for any other purpose 鈥 for the replication slave account.
Configure the slave to use the master. Using the MySQL
Monitor, this can be accomplished with the CHANGE
MASTER TO
statement:
mysqlS
>CHANGE MASTER TO
->MASTER_HOST='
->master_host
',MASTER_PORT=
->master_port
,MASTER_USER='
->slave_user
',MASTER_PASSWORD='
slave_password
';
where master_host
is the hostname
or IP address of the replication master,
master_port
is the port for the
slave to use for connecting to the master,
slave_user
is the username set up
for the slave on the master, and
slave_password
is the password
set for that user account in the previous step.
For example, to tell the slave to replicate from the MySQL
server whose hostname is
鈥rep-master
,鈥 using the
replication slave account created in the previous step, use
the following statement:
mysqlS
>CHANGE MASTER TO
->MASTER_HOST='rep-master'
->MASTER_PORT=3306,
->MASTER_USER='myslave'
->MASTER_PASSWORD='53cr37';
(For a complete list of clauses that can be used with this
statement, see Section聽13.6.2.1, 鈥CHANGE MASTER TO
Syntax鈥.)
You can also configure the slave to use the master by
setting the corresponding startup options in the slave
server's my.cnf
file. To configure the
slave in the same way as the preceding example
CHANGE MASTER TO
statement, the following
information would need to be included in the slave's
my.cnf
file:
[mysqld] master-host=rep-master master-port=3306 master-user=myslave master-password=53cr37
See Section聽6.1.3, 鈥淩eplication Options and Variables鈥, for additional
options that can be set in my.cnf
for
replication slaves.
Note: To provide
replication backup capability, you will also need to add an
ndb-connectstring
option to the slave's
my.cnf
file prior to starting the
replication process. See
Section聽15.10.9, 鈥淢ySQL Cluster Backups With Replication鈥, for
details.
If the master cluster is already in use, you can create a backup of the master and load this onto the slave to cut down on the amount of time required for the slave to synchronize itself with the master. If the slave is also running MySQL Cluster, this can be accomplished using the backup and restore procedure described in Section聽15.10.9, 鈥淢ySQL Cluster Backups With Replication鈥.
ndb-connectstring=management_host
[:port
]
In the event that you are not using MySQL Cluster on the replication slave, you can create a backup with this command on the replication master:
shellM
>mysqldump --master-data=1
Then import the resulting data dump onto the slave by
copying the dump file over to the slave. After this, you can
use the mysql client to import the data
from the dumpfile into the slave database as shown here,
where dump_file
is the name of
the file that was generated using
mysqldump on the master, and
db_name
is the name of the
database to be replicated:
shellS
>mysql -u root -p
db_name
<dump_file
For a complete list of options to use with mysqldump, see Section聽8.13, 鈥mysqldump 鈥 A Database Backup Program鈥.
Note that if you copy the data to the slave in this fashion,
you should make sure that the slave is started with the
--skip-slave-start
option on the
command line, or else include
skip-slave-start
in the slave's
my.cnf
file to keep it from trying to
connect to the master to begin replicating before all the
data has been loaded. Once the data loading has completed,
follow the additional steps outlined in the next two
sections.
Ensure that each MySQL server acting as a replication master
is configured with a unique server ID, and with binary
logging enabled, using the row format. (See
Section聽6.1.2, 鈥淩eplication Formats鈥.) These options can be
set either in the master server's
my.cnf
file, or on the command line
when starting the master mysqld process.
See Section聽15.10.6, 鈥淪tarting Replication (Single Replication Channel)鈥,
for information regarding the latter option.
This section outlines the procedure for starting MySQL CLuster replication using a single replication channel.
Start the MySQL replication master server by issuing this command:
shellM
>mysqld --nbdcluster --server-id=
id
\--log-bin --binlog-format=row &
where id
is this server's unique
ID (see
Section聽15.10.2, 鈥淎ssumptions and General Requirements鈥). This
starts the server's mysqld process with
binary logging enabled using the proper logging format.
Start the MySQL replication slave server as shown here:
shellS
>mysqld --ndbcluster --server-id=
id
&
where id
is the slave server's
unique ID. It is not necessary to enable logging on the
replication slave.
Note that you should use the
--skip-slave-start
option with this
command or else you should include
skip-slave-start
in the slave server's
my.cnf
file, unless you want
replication to begin immediately. With the use of this
option, the start of replication is delayed until the
appropriate START SLAVE
statement has
been issued, as explained in Step 4 below.
It is necessary to synchronize the slave server with the master server's replication binlog. If binary logging has not previously been running on the master, run the following statement on the slave:
mysqlS
>CHANGE MASTER TO
->MASTER_LOG_FILE='',
->MASTER_LOG_POS=4;
This instructs the slave to begin reading the master's
binary log from the log's starting point. Otherwise 鈥
that is, if you are loading data from the master using a
backup 鈥 see
Section聽15.10.8, 鈥淚mplementing Failover with MySQL Cluster鈥, for
information on how to obtain the correct values to use for
MASTER_LOG_FILE
and
MASTER_LOG_POS
in such cases.
Finally, you must instruct the slave to begin applying replication by issuing this command from the mysql client on the replication slave:
mysqlS
>START SLAVE;
This also initiates the transmission of replication data from the master to the slave.
It is also possible to use two replication channels, in a manner simlar to the procedure described in the next section; the differences between this and using a single replication channel are covered in Section聽15.10.7, 鈥淯sing Two Replication Channels鈥.
In a more complete example scenario, we envision two replication channels to provide redundancy and thereby guard against possible failure of a single replication channel. This requires a total of four replication servers, two masters for the master cluster and two slave servers for the slave cluster. For purposes of the discussion that follows, we assume that unique identifiers are assigned as shown here:
Server ID | Description |
1 | Master - primary replication channel (M) |
2 | Master - secondary replication channel (M') |
3 | Slave - primary replication channel (S) |
4 | Slave - secondary replication channel (S') |
Setting up replication with two channels is not radically
different from setting up a single replication channel. First,
the mysqld processes for the primary and
secondary replication masters must be started, followed by those
for the primary and secondary slaves. Then the replication
processes may be initiated by issuing the START
SLAVE
statement on each of the slaves. The commands
and the order in which they need to be issued are shown here:
Start the primary replication master:
shellM
>mysqld --ndbcluster --server-id=1 \
--log-bin --binlog-format=row &
Start the secondary replication master:
shellM'
>mysqld --ndbcluster --server-id=2 \
--log-bin --binlog-format=row &
Start the primary replication slave server:
shellS
>mysqld --ndbcluster --server-id=3 \
--skip-slave-start &
Start the secondary replication slave:
shellS'
>mysqld --ndbcluster --server-id=4 \
--skip-slave-start &
Finally, initiate replication on the primary channel by
executing the START SLAVE
statement on
the primary slave as shown here:
mysqlS
>START SLAVE;
As mentioned previously, it is not necessary to enable binary logging on replication slaves.
In the event that the primary Cluster replication process fails, it is possible to switch over to the secondary replication channel. The following procedure describes the steps required to accomplish this.
Obtain the time of the most recent global checkpoint (GCP).
That is, you need to determine the most recent epoch from
the ndb_apply_status
table on the slave
cluster, which can be found using the following query:
mysqlS'
>SELECT @latest:=MAX(epoch)
->FROM mysql.ndb_apply_status;
Using the information obtained from the query shown in Step
1, obtain the corresponding records from the
ndb_binlog_index
table on the master
cluster as shown here:
mysqlM'
>SELECT
->@file:=SUBSTRING_INDEX(File, '/', -1),
->@pos:=Position
->FROM mysql.ndb_binlog_index
->WHERE epoch > @latest
->ORDER BY epoch ASC LIMIT 1;
These are the records saved on the master since the failure
of the primary replication channel. We have employed a user
variable @latest
here to represent the
value obtained in Step 1. Of course, it is not possible for
one mysqld instance to access user
variables set on another server instance directly. These
values must be 鈥plugged in鈥 to the second query
manually or in application code.
Now it is possible to synchronize the secondary channel by running the following query on the secondary slave server:
mysqlS'
>CHANGE MASTER TO
->MASTER_LOG_FILE='@file',
->MASTER_LOG_POS=@pos;
Again we have employed user variables (in this case
@file
and @pos
) to
represent the values obtained in Step 2 and applied in Step
3; in practice these values must be inserted manually or
using application code that can access both of the servers
involved.
@file
is a string value such as
'/var/log/mysql/replication-master-bin.00001'
,
and so must be quoted when used in SQL or application
code. However, the value represented by
@pos
must not be
quoted. Although MySQL normally attempts to convert
strings to numbers, this case is an exception.
You can now initiate replication on the secondary channel by issuing the appropriate command on the secondary slave mysqld:
mysqlS'
>START SLAVE;
Once the secondary replication channel is active, you can investigate the failure of the primary and effect repairs. The precise actions required to do this will depend upon the reasons for which the primary channel failed.
If the failure is limited to a single server, it should (in
theory) be possible to replicate from
M
to S'
,
or from M'
to
S
; however, this has not yet been
tested.
This section discusses making backups and restoring from them using MySQL Cluster replication. We assume that the replication servers have already been configured as covered previously (see Section聽15.10.5, 鈥淧reparing the Cluster for Replication鈥, and the sections immediately following). This having been done, the procedure for making a backup and then restoring from it is as follows:
There are two different methods by which the backup may be started.
Method A:
This method requires that the cluster backup process was previously enabled on the master server, prior to starting the replication process. This can be done by including the line
ndb-connectstring=management_host
[:port
]
in a [MYSQL_CLUSTER]
section in the
my.cnf file
, where
management_host
is the IP
address or hostname of the NDB
management server for the master cluster, and
port
is the management
server's port number. Note that the port number needs to
be specified only if the default port (1186) is not
being used. (See
Section聽15.3.3, 鈥淢ulti-Computer Configuration鈥, for more
information about ports and port allocation in MySQL
Cluster.)
In this case, the backup can be started by executing this statement on the replication master:
shellM
>ndb_mgm -e "START BACKUP"
Method B:
If the my.cnf
file does not specify
where to find the management host, you can start the
backup process by passing this information to the
NDB
management client as part of the
START BACKUP
command, like this:
shellM
>ndb_mgm
management_host
:port
-e "START BACKUP"
where management_host
and
port
are the hostname and
port number of the management server. In our scenario as
outlined earlier (see
Section聽15.10.5, 鈥淧reparing the Cluster for Replication鈥),
this would be executed as follows:
shellM
>ndb_mgm rep-master:1186 -e "START BACKUP"
In either case, it is highly advisable to allow any pending transactions to be completed before beginning the backup, and then not to permit any new transactions to begin during the backup process.
Copy the cluster backup files to the slave that is being
brought on line. Each system running an
ndbd process for the master cluster will
have cluster backup files located on it, and
all of these files must be copied to
the slave to ensure a successful restore. The backup files
can be copied into any directory on the computer where the
slave management host resides, so long as the MySQL and NDB
binaries have read permissions in that directory. In this
case, we will assume that these files have been copied into
the directory /var/BACKUPS/BACKUP-1
.
It is not necessary that the slave cluster have the same
number of ndbd processes (data nodes) as
the master; however, it is highly recommended this number be
the same. It is necessary that the
slave be started with the
--skip-slave-start
option, to prevent
premature startup of the replication process.
Create any databases on the slave cluster that are present
on the master cluster that are to be replicated to the
slave. Important: A
CREATE SCHEMA
statement corresponding to
each database to be replicated must be executed on each data
node in the slave cluster.
Reset the slave cluster using this statement in the MySQL Monitor:
mysqlS
>RESET SLAVE;
It is important to make sure that the slave's
apply_status
table does not contain any
records prior to running the restore process. You can
accomplish this by running this SQL statement on the slave:
mysqlS
>DELETE FROM mysql.ndb_apply_status;
You can now start the cluster restoration process on the
replication slave using the ndb_restore
command for each backup file in turn. For the first of
these, it is necessary to include the -m
option to restore the cluster metadata:
shellS
>ndb_restore -c
slave_host
:port
-nnode-id
\-b
backup-id
-m -rdir
dir
is the path to the directory
where the backup files have been placed on the replication
slave. For the ndb_restore commands
corresponding to the remaining backup files, the
-m
option should not
be used.
For restoring from a master cluster with four data nodes (as
shown in the figure in
Section聽15.10, 鈥淢ySQL Cluster Replication鈥) where the
backup files have been copied to the directory
/var/BACKUPS/BACKUP-1
, the proper
sequence of commands to be executed on the slave might look
like this:
shellS
>ndb_restore -c rep-slave:1186 -n 2 -b 1 -m \
-r ./var/BACKUPS/BACKUP-1
shellS
>ndb_restore -c rep-slave:1186 -n 3 -b 1 \
-r ./var/BACKUPS/BACKUP-1
shellS
>ndb_restore -c rep-slave:1186 -n 4 -b 1 \
-r ./var/BACKUPS/BACKUP-1
shellS
>ndb_restore -c rep-slave:1186 -n 5 -b 1 -e \
-r ./var/BACKUPS/BACKUP-1
The -e
(or
--restore-epoch
) option in the final
invocation of ndb_restore in this
example is required in order that the epoch is written to
the slave mysql.ndb_apply_status
.
Without this information, the slave will not be able to
synchronize properly with the master. (See
Section聽15.8.3, 鈥ndb_restore 鈥 Restore a Cluster Backup鈥.)
Now you need to obtain the most recent epoch from the
ndb_apply_status
table on the slave (as
discussed in
Section聽15.10.8, 鈥淚mplementing Failover with MySQL Cluster鈥):
mysqlS
>SELECT @latest:=MAX(epoch)
FROM mysql.ndb_apply_status;
Using @latest
as the epoch value obtained
in the previous step, you can obtain the correct starting
position @pos
in the correct binary log
file @file
from the master's
mysql.ndb_binlog_index
table using the
query shown here:
mysqlM
>SELECT
->@file:=SUBSTRING_INDEX(File, '/', -1),
->@pos:=Position
->FROM mysql.ndb_binlog_index
->WHERE epoch > @latest
->ORDER BY epoch ASC LIMIT 1;
Using the values obtained in the previous step, you can now
issue the appropriate CHANGE MASTER TO
statement in the slave's mysql client:
mysqlS
>CHANGE MASTER TO
->MASTER_LOG_FILE='@file',
->MASTER_LOG_POS=@pos;
Now that the slave 鈥knows鈥 from what point in
which binlog
file to start reading data
from the master, you can cause the slave to begin
replicating with this standard MySQL statement:
mysqlS
>START SLAVE;
To perform a backup and restore on a second replication channel, it is necessary only to repeat these steps, substituting the hostnames and IDs of the secondary master and slave for those of the primary master and slave replication servers where appropriate, and running the preceding statements on them.
For additional information on performing Cluster backups and restoring Cluster from backups, see Section聽15.8, 鈥淥n-line Backup of MySQL Cluster鈥.
It is possible to automate much of the process described in
the previous section (see
Section聽15.10.9, 鈥淢ySQL Cluster Backups With Replication鈥). The
following Perl script reset-slave.pl
serves as an example of how you can do this.
#!/user/bin/perl -w # file: reset-slave.pl # Copyright 漏2005 MySQL AB # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 of the License, or # (at your option) any later version. # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # You should have received a copy of the GNU General Public License # along with this program; if not, write to: # Free Software Foundation, Inc. # 59 Temple Place, Suite 330 # Boston, MA 02111-1307 USA # # Version 1.1 ######################## Includes ############################### use DBI; ######################## Globals ################################ my $m_host=''; my $m_port=''; my $m_user=''; my $m_pass=''; my $s_host=''; my $s_port=''; my $s_user=''; my $s_pass=''; my $dbhM=''; my $dbhS=''; ####################### Sub Prototypes ########################## sub CollectCommandPromptInfo; sub ConnectToDatabases; sub DisconnectFromDatabases; sub GetSlaveEpoch; sub GetMasterInfo; sub UpdateSlave; ######################## Program Main ########################### CollectCommandPromptInfo; ConnectToDatabases; GetSlaveEpoch; GetMasterInfo; UpdateSlave; DisconnectFromDatabases; ################## Collect Command Prompt Info ################## sub CollectCommandPromptInfo { ### Check that user has supplied correct number of command line args die "Usage:\n reset-slave >master MySQL host< >master MySQL port< \n >master user< >master pass< >slave MySQL host< \n >slave MySQL port< >slave user< >slave pass< \n All 8 arguments must be passed. Use BLANK for NULL passwords\n" unless @ARGV == 8; $m_host = $ARGV[0]; $m_port = $ARGV[1]; $m_user = $ARGV[2]; $m_pass = $ARGV[3]; $s_host = $ARGV[4]; $s_port = $ARGV[5]; $s_user = $ARGV[6]; $s_pass = $ARGV[7]; if ($m_pass eq "BLANK") { $m_pass = '';} if ($s_pass eq "BLANK") { $s_pass = '';} } ############### Make connections to both databases ############# sub ConnectToDatabases { ### Connect to both master and slave cluster databases ### Connect to master $dbhM = DBI->connect( "dbi:mysql:database=mysql;host=$m_host;port=$m_port", "$m_user", "$m_pass") or die "Can't connect to Master Cluster MySQL process! Error: $DBI::errstr\n"; ### Connect to slave $dbhS = DBI->connect( "dbi:mysql:database=mysql;host=$s_host", "$s_user", "$s_pass") or die "Can't connect to Slave Cluster MySQL process! Error: $DBI::errstr\n"; } ################ Disconnect from both databases ################ sub DisconnectFromDatabases { ### Disconnect from master $dbhM->disconnect or warn " Disconnection failed: $DBI::errstr\n"; ### Disconnect from slave $dbhS->disconnect or warn " Disconnection failed: $DBI::errstr\n"; } ###################### Find the last good GCI ################## sub GetSlaveEpoch { $sth = $dbhS->prepare("SELECT MAX(epoch) FROM mysql.ndb_apply_status;") or die "Error while preparing to select epoch from slave: ", $dbhS->errstr; $sth->execute or die "Selecting epoch from slave error: ", $sth->errstr; $sth->bind_col (1, \$epoch); $sth->fetch; print "\tSlave Epoch = $epoch\n"; $sth->finish; } ####### Find the position of the last GCI in the binlog ######## sub GetMasterInfo { $sth = $dbhM->prepare("SELECT SUBSTRING_INDEX(File, '/', -1), Position FROM mysql.ndb_binlog_index WHERE epoch > $epoch ORDER BY epoch ASC LIMIT 1;") or die "Prepare to select from master error: ", $dbhM->errstr; $sth->execute or die "Selecting from master error: ", $sth->errstr; $sth->bind_col (1, \$binlog); $sth->bind_col (2, \$binpos); $sth->fetch; print "\tMaster bin log = $binlog\n"; print "\tMaster Bin Log position = $binpos\n"; $sth->finish; } ########## Set the slave to process from that location ######### sub UpdateSlave { $sth = $dbhS->prepare("CHANGE MASTER TO MASTER_LOG_FILE='$binlog', MASTER_LOG_POS=$binpos;") or die "Prepare to CHANGE MASTER error: ", $dbhS->errstr; $sth->execute or die "CHNAGE MASTER on slave error: ", $sth->errstr; $sth->finish; print "\tSlave has been updated. You may now start the slave.\n"; } # end reset-slave.pl
Beginning with MySQL 5.1.6, it is possible to store the
non-indexed columns of NDB
tables on disk,
rather than in RAM as with previous versions of MySQL Cluster.
As part of implementing Cluster Disk Data work, a number of improvements were made in MySQL Cluster for the efficient handling of very large amounts (terabytes) of data during node recovery and restart. These include a 鈥no-steal鈥 algorithm for synchronising a starting node with very large data sets. For more information, see the paper Recovery Principles of MySQL Cluster 5.1, by MySQL Cluster developers Mikael Ronstr枚m and Jonas Oreland.
This section discusses Disk Data objects 鈥 which include tables, log file groups, and tablespaces 鈥 as well as how to create and drop them.
Assuming that you have already set up a MySQL Cluster with all nodes (including management and SQL nodes) running MySQL 5.1.6 or newer, the basic steps for creating a Cluster table on disk are as follows:
Create a log file group, and assign one or more undo log files to it (an undo log file is also referred as an undofile).
In MySQL 5.1, undo log files are necessary only for Disk Data tables. They are no longer used for tables that are stored in memory.
Create a tablespace, and assign the log file group to it, as well as one or more data files.
Create a Disk Data table that uses this tablespace for data storage.
Each of these tasks can be accomplished using SQL statements, as shown in the following example.
We create a log file group named lg_1
using CREATE LOGFILE GROUP
. This log file
group is to be made up of two undo log files, which we name
undo_1.dat
and
undo_2.dat
, whose initial sizes are 16
MB and 12 MB, respectively. (The default initial size for an
undo log file is 128 MB.) Optionally, you can also specify a
size for the log file group's UNDO
buffer, or allow it to assume the default value of 8 MB. In
this example, we set the UNDO buffer's size at 2 MB. A log
file group must be created with an undo log file; so we add
undo_1.dat
to lg_1
in this CREATE LOGFILE GROUP
statement:
CREATE LOGFILE GROUP lg_1 ADD UNDOFILE 'undo_1.dat' INITIAL_SIZE 16M UNDO_BUFFER_SIZE 2M ENGINE NDB;
To add undo_2.dat
to the log file
group, use the following ALTER LOGFILE
GROUP
statement:
ALTER LOGFILE GROUP lg_1 ADD UNDOFILE 'undo_2.dat' INITIAL_SIZE 12M ENGINE NDB;
Some items of note:
The .dat
file extension used here
is not required. We use it merely to make the log and
data files easily recognisable.
Every CREATE LOGFILE GROUP
and
ALTER LOGFILE GROUP
statement must
include an ENGINE
clause. In MySQL
5.1, the permitted values for this clause are
NDB
and
NDBCLUSTER
.
In MySQL 5.1.8 and later, there can exist only one log file group at any given time.
When you add an undo log file to a log file group using
ADD UNDOFILE
'
, a file
with the name filename
'filename
is
created in the
ndb_
directory within the nodeid
_fsDataDirectory
of
each data node in the cluster, where
nodeid
is the node ID of the
data node.
UNDO_BUFFER_SIZE
is limited by the
amount of system memory available.
For more information about the CREATE LOGFILE
GROUP
statement, see
Section聽13.1.9, 鈥CREATE LOGFILE GROUP
Syntax鈥. For more
information about ALTER LOGFILE
GROUP
, see
Section聽13.1.3, 鈥ALTER LOGFILE GROUP
Syntax鈥.
Now we can create a tablespace, which contains files to be used by MySQL Cluster Disk Data tables for storing their data. A tablespace is also associated with a particular log file group. When creating a new tablespace, you must specify the log file group which it is to use for undo logging; you must also specify a data file. You can add more data files to the tablespace after it the tablespace is created; it is also possible to drop data files from a tablespace (an example of dropping data files is provided later in this section).
Assume that we wish to create a tablespace named
ts_1
which uses lg_1
as its log file group. This tablespace is to contain two
data files named data_1.dat
and
data_2.dat
, whose initial sizes are 32
MB and 48 MB, respectively. (The default value for
INITIAL_SIZE
is 128 MB.) We can do this
using two SQL statements, as shown here:
CREATE TABLESPACE ts_1 ADD DATAFILE 'data_1.dat' USE LOGFILE GROUP lg_1 INITIAL_SIZE 32M ENGINE NDB; ALTER TABLESPACE ts_1 ADD DATAFILE 'data_2.dat' INITIAL_SIZE 48M ENGINE NDB;
The CREATE TABLESPACE
statement creates a
tablespace ts_1
with the data file
data_1.dat
, and associates
ts_1
with log file group
lg_1
. The ALTER
TABLESPACE
adds the second data file
(data_2.dat
).
Some items of note:
As is the case with the filenames used here for undo log
files, there is no special significance for the
.dat
file extension; it is used
merely for easy recognition.
All CREATE TABLESPACE
and
ALTER TABLESPACE
statements must
contain an ENGINE
clause; only tables
using the same storage engine as the tablespace can be
created in the tablespace. In MySQL 5.1, the only
permitted values for this clause are
NDB
and
NDBCLUSTER
.
For more information about the CREATE
TABLESPACE
and ALTER
TABLESPACE
statements, see
Section聽13.1.10, 鈥CREATE TABLESPACE
Syntax鈥, and
Section聽13.1.4, 鈥ALTER TABLESPACE
Syntax鈥.
Now it is possible to create a table whose non-indexed
columns are stored on disk in the tablespace
ts_1
:
CREATE TABLE dt_1 ( member_id INT UNSIGNED NOT NULL AUTO_INCREMENT PRIMARY KEY, last_name VARCHAR(50) NOT NULL, first_name VARCHAR(50) NOT NULL, dob DATE NOT NULL, joined DATE NOT NULL, INDEX(last_name, first_name) ) TABLESPACE ts_1 STORAGE DISK ENGINE NDB;
The TABLESPACE ... STORAGE DISK
clause
tells the NDB Cluster
storage engine to
use tablespace ts_1
for disk data
storage.
Once table ts_1
has been created as
shown, you can perform INSERT
,
SELECT
, UPDATE
, and
DELETE
statements on it just as you would
with any other MySQL table.
For table dt_1
as it has been defined
here, only the dob
and
joined
columns are stored on disk. This
is because there are indexes on the id
,
last_name
, and
first_name
columns, and so data belonging
to these columns is stored in RAM. In MySQL 5.1, only
non-indexed columns can be held on disk; indexes and indexed
column data continue to be stored in memory. This trade-off
between the use of indexes and conservation of RAM is
something you must keep in mind as you design Disk Data
tables.
Performance note.聽 The performance of a cluster using Disk Data storage is greatly improved if Disk Data files are kept on a separate physical disk from the data node filesystem. This must be done for each data node in the cluster to derive any noticeable benefit.
You may use absolute and relative filesystem paths with
ADD UNDOFILE
and ADD
DATAFILE
. Relative paths are calculated relative to
the data node's data directory.
A log file group, a tablespace, and any Disk Data tables using these must be created in a particular order. The same is true for dropping any of these objects:
A log file group cannot be dropped, so long as any tablespaces are using it.
A tablespace cannot be dropped as long as it contains any data files.
You cannot drop any data files from a tablespace as long as there remain any tables which are using the tablespace.
Beginning with MySQL 5.1.12, it is no longer possible to drop files created in association with a different tablespace than the one with which the files were created. (Bug#20053)
For example, to drop all the objects created so far in this section, you would use the following statements:
mysql>DROP TABLE dt_1;
mysql>ALTER TABLESPACE ts_1
->DROP DATAFILE 'data_2.dat'
->ENGINE NDB;
mysql>ALTER TABLESPACE ts_1
->DROP DATAFILE 'data_1.dat'
->ENGINE NDB;
mysql>DROP TABLESPACE ts_1
->ENGINE NDB;
mysql>DROP LOGFILE GROUP lg_1
->ENGINE NDB;
These statements must be performed in the order shown, except
that the two ALTER TABLESPACE ... DROP
DATAFILE
statements may be executed in either order.
You can obtain information about data files used by Disk Data
tables by querying the FILES
table in the
INFORMATION_SCHEMA
database. An extra
鈥NULL
row鈥 was added to this
table in MySQL 5.1.14 for providing additional information about
undo log files. For more information and examples of use, see
Section聽22.21, 鈥淭he INFORMATION_SCHEMA FILES
Table鈥.
The following items apply to Disk Data storage requirements:
Variable-length columns of Disk Data tables take up a fixed amount of space. For each row, this is equal to the space required to store the largest possible value for that column.
For general information about calculating these values, see Section聽11.5, 鈥淒ata Type Storage Requirements鈥.
In a Disk Data table, the first 256 bytes of a
TEXT
or BLOB
column
are stored in memory; only the remainder is stored on
disk.
Starting the cluster with the --initial
option does not remove Disk Data files.
You must remove these manually prior to performing an initial
restart of the cluster.
Configuration parameters affecting Disk Data behaviour include the following:
This determines the amount of space used for caching pages
on disk, and is set in the [NDBD]
or
[NDBD DEFAULT]
section of the
config.ini
file. It is measured in
bytes. Each page takes up 32 KB. This means that Cluster
Disk Data storage always uses N
*
32 KB memory where N
is some
non-negative integer.
This determines the amount of memory that is used for log
buffers, disk operations (such as page requests and wait
queues), and metadata for tablespaces, log file groups,
UNDO
files, and data files. It can be set
in the [NDBD]
or [NDBD
DEFAULT]
section of the
config.ini
configuration file, and is
measured in bytes.
The default value is 20M
.
The OPTIMIZE TABLE
statement does not have
any effect on Disk Data tables.
Even before design of NDB Cluster
began in
1996, it was evident that one of the major problems to be
encountered in building parallel databases would be communication
between the nodes in the network. For this reason, NDB
Cluster
was designed from the very beginning to allow
for the use of a number of different data transport mechanisms. In
this Manual, we use the term transporter
for these.
The MySQL Cluster codebase includes support for four different transporters:
TCP/IP using 100 Mbps or gigabit Ethernet, as discussed in Section聽15.4.4.7, 鈥淐luster TCP/IP Connections鈥.
Direct (machine-to-machine) TCP/IP; although this transporter uses the same TCP/IP protocol as mentioned in the previous item, it requires setting up the hardware differently and is configured differently as well. For this reason, it is considered a separate transport mechanism for MySQL Cluster. See Section聽15.4.4.8, 鈥淭CP/IP Connections Using Direct Connections鈥, for details.
Shared memory (SHM). For more information about SHM, see Section聽15.4.4.9, 鈥淪hared-Memory Connections鈥.
Scalable Coherent Interface (SCI), as described in the next section of this chapter, Section聽15.4.4.10, 鈥淪CI Transport Connections鈥.
Most users today employ TCP/IP over Ethernet because it is ubiquitous. TCP/IP is also by far the best-tested transporter for use with MySQL Cluster.
We are working to make sure that communication with the ndbd process is made in 鈥chunks鈥 that are as large as possible because this benefits all types of data transmission.
For users who desire it, it is also possible to use cluster interconnects to enhance performance even further. There are two ways to achieve this: Either a custom transporter can be designed to handle this case, or you can use socket implementations that bypass the TCP/IP stack to one extent or another. We have experimented with both of these techniques using the SCI (Scalable Coherent Interface) technology developed by Dolphin.
In this section, we show how to adapt a cluster configured for normal TCP/IP communication to use SCI Sockets instead. This documentation is based on SCI Sockets version 2.3.0 as of 01 October 2004.
Prerequisites
Any machines with which you wish to use SCI Sockets must be equipped with SCI cards.
It is possible to use SCI Sockets with any version of MySQL Cluster. No special builds are needed because it uses normal socket calls which are already available in MySQL Cluster. However, SCI Sockets are currently supported only on the Linux 2.4 and 2.6 kernels. SCI Transporters have been tested successfully on additional operating systems although we have verified these only with Linux 2.4 to date.
There are essentially four requirements for SCI Sockets:
Building the SCI Socket libraries.
Installation of the SCI Socket kernel libraries.
Installation of one or two configuration files.
The SCI Socket kernel library must enabled either for the entire machine or for the shell where the MySQL Cluster processes are started.
This process needs to be repeated for each machine in the cluster where you plan to use SCI Sockets for inter-node communication.
Two packages need to be retrieved to get SCI Sockets working:
The source code package containing the DIS support libraries for the SCI Sockets libraries.
The source code package for the SCI Socket libraries themselves.
Currently, these are available only in source code format. The
latest versions of these packages at the time of this writing
were available as (respectively)
DIS_GPL_2_5_0_SEP_10_2004.tar.gz
and
SCI_SOCKET_2_3_0_OKT_01_2004.tar.gz
. You
should be able to find these (or possibly newer versions) at
http://www.dolphinics.no/support/downloads.html.
Package Installation
Once you have obtained the library packages, the next step is to unpack them into appropriate directories, with the SCI Sockets library unpacked into a directory below the DIS code. Next, you need to build the libraries. This example shows the commands used on Linux/x86 to perform this task:
shell>tar xzf DIS_GPL_2_5_0_SEP_10_2004.tar.gz
shell>cd DIS_GPL_2_5_0_SEP_10_2004/src/
shell>tar xzf ../../SCI_SOCKET_2_3_0_OKT_01_2004.tar.gz
shell>cd ../adm/bin/Linux_pkgs
shell>./make_PSB_66_release
It is possible to build these libraries for some 64-bit procesors. To build the libraries for Opteron CPUs using the 64-bit extensions, run make_PSB_66_X86_64_release rather than make_PSB_66_release. If the build is made on an Itanium machine, you should use make_PSB_66_IA64_release. The X86-64 variant should work for Intel EM64T architectures but this has not yet (to our knowledge) been tested.
Once the build process is complete, the compiled libraries will
be found in a zipped tar file with a name along the lines of
DIS-
.
It is now time to install the package in the proper place. In
this example we will place the installation in
<operating-system>
-time
-date
/opt/DIS
.
(Note: You will most likely
need to run the following as the system root
user.)
shell>cp DIS_Linux_2.4.20-8_181004.tar.gz /opt/
shell>cd /opt
shell>tar xzf DIS_Linux_2.4.20-8_181004.tar.gz
shell>mv DIS_Linux_2.4.20-8_181004 DIS
Network Configuration
Now that all the libraries and binaries are in their proper place, we need to ensure that the SCI cards have proper node IDs within the SCI address space.
It is also necessary to decide on the network structure before proceeding. There are three types of network structures which can be used in this context:
A simple one-dimensional ring
One or more SCI switches with one ring per switch port
A two- or three-dimensional torus.
Each of these topologies has its own method for providing node IDs. We discuss each of them in brief.
A simple ring uses node IDs which are non-zero multiples of 4: 4, 8, 12,...
The next possibility uses SCI switches. An SCI switch has 8 ports, each of which can support a ring. It is necessary to make sure that different rings use different node ID spaces. In a typical configuration, the first port uses node IDs below 64 (4 鈥 60), the next 64 node IDs (68 鈥 124) are assigned to the next port, and so on, with node IDs 452 鈥 508 being assigned to the eighth port.
Two- and three-dimensional torus network structures take into account where each node is located in each dimension, incrementing by 4 for each node in the first dimension, by 64 in the second dimension, and (where applicable) by 1024 in the third dimension. See Dolphin's Web site for more thorough documentation.
In our testing we have used switches, although most large cluster installations use 2- or 3-dimensional torus structures. The advantage provided by switches is that, with dual SCI cards and dual switches, it is possible to build with relative ease a redundant network where the average failover time on the SCI network is on the order of 100 microseconds. This is supported by the SCI transporter in MySQL Cluster and is also under development for the SCI Socket implementation.
Failover for the 2D/3D torus is also possible but requires sending out new routing indexes to all nodes. However, this requires only 100 milliseconds or so to complete and should be acceptable for most high-availability cases.
By placing cluster data nodes properly within the switched architecture, it is possible to use 2 switches to build a structure whereby 16 computers can be interconnected and no single failure can hinder more than one of them. With 32 computers and 2 switches it is possible to configure the cluster in such a manner that no single failure can cause the loss of more than two nodes; in this case, it is also possible to know which pair of nodes is affected. Thus, by placing the two nodes in separate node groups, it is possible to build a 鈥safe鈥 MySQL Cluster installation.
To set the node ID for an SCI card use the following command in
the /opt/DIS/sbin
directory. In this
example, -c 1
refers to the number of the SCI
card (this is always 1 if there is only 1 card in the machine);
-a 0
refers to adapter 0; and
68
is the node ID:
shell> ./sciconfig -c 1 -a 0 -n 68
If you have multiple SCI cards in the same machine, you can
determine which card has which slot by issuing the following
command (again we assume that the current working directory is
/opt/DIS/sbin
):
shell> ./sciconfig -c 1 -gsn
This will give you the SCI card's serial number. Then repeat
this procedure with -c 2
, and so on, for each
card in the machine. Once you have matched each card with a
slot, you can set node IDs for all cards.
After the necessary libraries and binaries are installed, and
the SCI node IDs are set, the next step is to set up the mapping
from hostnames (or IP addresses) to SCI node IDs. This is done
in the SCI sockets configuration file, which should be saved as
/etc/sci/scisock.conf
. In this file, each
SCI node ID is mapped through the proper SCI card to the
hostname or IP address that it is to communicate with. Here is a
very simple example of such a configuration file:
#host #nodeId alpha 8 beta 12 192.168.10.20 16
It is also possible to limit the configuration so that it
applies only to a subset of the available ports for these hosts.
An additional configuration file
/etc/sci/scisock_opt.conf
can be used to
accomplish this, as shown here:
#-key -type -values EnablePortsByDefault yes EnablePort tcp 2200 DisablePort tcp 2201 EnablePortRange tcp 2202 2219 DisablePortRange tcp 2220 2231
Driver Installation
With the configuration files in place, the drivers can be installed.
First, the low-level drivers and then the SCI socket driver need to be installed:
shell>cd DIS/sbin/
shell>./drv-install add PSB66
shell>./scisocket-install add
If desired, the installation can be checked by invoking a script which verifies that all nodes in the SCI socket configuration files are accessible:
shell>cd /opt/DIS/sbin/
shell>./status.sh
If you discover an error and need to change the SCI socket configuration, it is necessary to use ksocketconfig to accomplish this task:
shell>cd /opt/DIS/util
shell>./ksocketconfig -f
Testing the Setup
To ensure that SCI sockets are actually being used, you can
employ the latency_bench test program. Using
this utility's server component, clients can connect to the
server to test the latency of the connection. Determining
whether SCI is enabled should be fairly simple from observing
the latency. (Note: Before
using latency_bench, it is necessary to set
the LD_PRELOAD
environment variable as shown
later in this section.)
To set up a server, use the following:
shell>cd /opt/DIS/bin/socket
shell>./latency_bench -server
To run a client, use latency_bench again,
except this time with the -client
option:
shell>cd /opt/DIS/bin/socket
shell>./latency_bench -client
server_hostname
SCI socket configuration should now be complete and MySQL Cluster ready to use both SCI Sockets and the SCI transporter (see Section聽15.4.4.10, 鈥淪CI Transport Connections鈥).
Starting the Cluster
The next step in the process is to start MySQL Cluster. To
enable usage of SCI Sockets it is necessary to set the
environment variable LD_PRELOAD
before
starting ndbd, mysqld, and
ndb_mgmd. This variable should point to the
kernel library for SCI Sockets.
To start ndbd in a bash shell, do the following:
bash-shell>export LD_PRELOAD=/opt/DIS/lib/libkscisock.so
bash-shell>ndbd
In a tcsh environment the same thing can be accomplished with:
tcsh-shell>setenv LD_PRELOAD=/opt/DIS/lib/libkscisock.so
tcsh-shell>ndbd
Note: MySQL Cluster can use only the kernel variant of SCI Sockets.
The ndbd process has a number of simple constructs which are used to access the data in a MySQL Cluster. We have created a very simple benchmark to check the performance of each of these and the effects which various interconnects have on their performance.
There are four access methods:
Primary key access
This is access of a record through its primary key. In the simplest case, only one record is accessed at a time, which means that the full cost of setting up a number of TCP/IP messages and a number of costs for context switching are borne by this single request. In the case where multiple primary key accesses are sent in one batch, those accesses share the cost of setting up the necessary TCP/IP messages and context switches. If the TCP/IP messages are for different destinations, additional TCP/IP messages need to be set up.
Unique key access
Unique key accesses are similar to primary key accesses, except that a unique key access is executed as a read on an index table followed by a primary key access on the table. However, only one request is sent from the MySQL Server, and the read of the index table is handled by ndbd. Such requests also benefit from batching.
Full table scan
When no indexes exist for a lookup on a table, a full table scan is performed. This is sent as a single request to the ndbd process, which then divides the table scan into a set of parallel scans on all cluster ndbd processes. In future versions of MySQL Cluster, an SQL node will be able to filter some of these scans.
Range scan using ordered index
When an ordered index is used, it performs a scan in the same manner as the full table scan, except that it scans only those records which are in the range used by the query transmitted by the MySQL server (SQL node). All partitions are scanned in parallel when all bound index attributes include all attributes in the partitioning key.
To check the base performance of these access methods, we have developed a set of benchmarks. One such benchmark, testReadPerf, tests simple and batched primary and unique key accesses. This benchmark also measures the setup cost of range scans by issuing scans returning a single record. There is also a variant of this benchmark which uses a range scan to fetch a batch of records.
In this way, we can determine the cost of both a single key access and a single record scan access, as well as measure the impact of the communication media used, on base access methods.
In our tests, we ran the base benchmarks for both a normal transporter using TCP/IP sockets and a similar setup using SCI sockets. The figures reported in the following table are for small accesses of 20 records per access. The difference between serial and batched access decreases by a factor of 3 to 4 when using 2KB records instead. SCI Sockets were not tested with 2KB records. Tests were performed on a cluster with 2 data nodes running on 2 dual-CPU machines equipped with AMD MP1900+ processors.
Access Type | TCP/IP Sockets | SCI Socket |
Serial pk access | 400 microseconds | 160 microseconds |
Batched pk access | 28 microseconds | 22 microseconds |
Serial uk access | 500 microseconds | 250 microseconds |
Batched uk access | 70 microseconds | 36 microseconds |
Indexed eq-bound | 1250 microseconds | 750 microseconds |
Index range | 24 microseconds | 12 microseconds |
We also performed another set of tests to check the performance of SCI Sockets vis-脿-vis that of the SCI transporter, and both of these as compared with the TCP/IP transporter. All these tests used primary key accesses either serially and multi-threaded, or multi-threaded and batched.
The tests showed that SCI sockets were about 100% faster than TCP/IP. The SCI transporter was faster in most cases compared to SCI sockets. One notable case occurred with many threads in the test program, which showed that the SCI transporter did not perform very well when used for the mysqld process.
Our overall conclusion was that, for most benchmarks, using SCI sockets improves performance by approximately 100% over TCP/IP, except in rare instances when communication performance is not an issue. This can occur when scan filters make up most of processing time or when very large batches of primary key accesses are achieved. In that case, the CPU processing in the ndbd processes becomes a fairly large part of the overhead.
Using the SCI transporter instead of SCI Sockets is only of interest in communicating between ndbd processes. Using the SCI transporter is also only of interest if a CPU can be dedicated to the ndbd process because the SCI transporter ensures that this process will never go to sleep. It is also important to ensure that the ndbd process priority is set in such a way that the process does not lose priority due to running for an extended period of time, as can be done by locking processes to CPUs in Linux 2.6. If such a configuration is possible, the ndbd process will benefit by 10鈥70% as compared with using SCI sockets. (The larger figures will be seen when performing updates and probably on parallel scan operations as well.)
There are several other optimized socket implementations for computer clusters, including Myrinet, Gigabit Ethernet, Infiniband and the VIA interface. We have tested MySQL Cluster so far only with SCI sockets. See Section聽15.12.1, 鈥淐onfiguring MySQL Cluster to use SCI Sockets鈥 for information on how to set up SCI sockets using ordinary TCP/IP for MySQL Cluster.
In this section, we provide a list of known limitations in MySQL
Cluster releases in the 5.1.x series compared to
features available when using the MyISAM
and
InnoDB
storage engines. Currently, there are no
plans to address these in coming releases of MySQL
5.1; however, we will attempt to supply fixes for
these issues in subsequent release series. If you check the
鈥Cluster鈥 category in the MySQL bugs database at
http://bugs.mysql.com, you can find known bugs
which (if marked 鈥5.1鈥) we intend to
correct in upcoming releases of MySQL 5.1.
The list here is intended to be complete with respect to the conditions just set forth. You can report any discrepancies that you encounter to the MySQL bugs database using the instructions given in Section聽1.8, 鈥淗ow to Report Bugs or Problems鈥. If we do not plan to fix the problem in MySQL 5.1, we will add it to the list.
(Note: See the end of this section for a list of issues in MySQL 5.0 Cluster that have been resolved in the current version.)
Limitations and other issues specific to MySQL Cluster Replication are described in Section聽15.10.3, 鈥淜nown Issues鈥.
Noncompliance in syntax (resulting in errors when running existing applications):
Temporary tables are not supported.
You cannot create indexes on NDB
table
columns that use any of the TEXT
or
BLOB
data types.
The NDB
storage engine also does not
support FULLTEXT
indexes (these are
supported by MyISAM
only).
However, you can index VARCHAR
columns
of NDB
tables.
A BIT
column cannot be a primary key,
unique key, or index, nor can it be part of a composite
primary key, unique key, or index.
Geometry datatypes (WKT
and
WKB
) are supported in
NDB
tables in MySQL 5.1.
However, spatial indexes are not supported.
CREATE TABLE
statements may be no more
than 4096 characters in length. This limitation
affects MySQL 5.1.6, 5.1.7, and 5.1.8 only.
(See Bug#17813)
In MySQL 5.1.7 and earlier, INSERT
IGNORE
, UPDATE IGNORE
, and
REPLACE
are supported only for primary
keys, but not for unique keys. One possible workaround is
to remove the constraint by dropping the unique index,
perform any inserts, and then add the unique index again.
This limitation is removed for INSERT
IGNORE
and REPLACE
in MySQL
5.1.8. (Bug#17431)
Support for user-defined partitioning for MySQL Cluster in
MySQL 5.1 is restricted to
[LINEAR
] KEY
partitioning. Beginning with MySQL 5.1.12, using any other
partitioning type with ENGINE=NDB
or
ENGINE=NDBCLUSTER
in a CREATE
TABLE
statement results in an error.
As of MySQL 5.1.6, all Cluster tables are by default
partitioned by KEY
using the table's
primary key as the partitioning key. If no primary key is
explicitly set for the table, the 鈥hidden鈥
primary key automatically created by the
NDB
storage engine is used instead. For
additional discussion of these and related issues, see
Section聽16.2.4, 鈥KEY
Partitioning鈥.
It is not possible to drop partitions from
NDB
tables using ALTER TABLE
... DROP PARTITION
. The other partitioning
extensions to ALTER TABLE
鈥
ADD PARTITION
, REORGANIZE
PARTITION
, and COALESCE
PARTITION
鈥 are supported for Cluster
tables, but use copying and so are not optimised. See
Section聽16.3.1, 鈥淢anagement of RANGE
and LIST
Partitions鈥 and
Section聽13.1.2, 鈥ALTER TABLE
Syntax鈥.
When using row-based replication with MySQL Cluster,
binary logging cannot be disabled. That is, the
NDB
storage engine ignores the value of
SQL_LOG_BIN
. (Bug#16680)
The auto_increment_increment
and
auto_increment_offset
server system
variables are not supported for Cluster replication.
Non-compliance in limits or behavior (may result in errors when running existing applications):
Memory Usage:
Memory comsumed when data is inserted into an
NDB
table is not automatically
recovered when deleted, as it is with other storage
engines. Instead, the following rules hold true:
A DELETE
statement on an
NDB
table makes the memory formerly
used by the deleted rows available for re-use by
inserts on the same table only. The memory cannot be
used by other NDB
tables.
A DROP TABLE
or
TRUNCATE
operation on an
NDB
table frees the memory that was
used by this table for re-use by any
NDB
table 鈥 either by the
same table or by another NDB
table.
(Recall that TRUNCATE
drops and
re-creates the table. See Section聽13.2.9, 鈥TRUNCATE
Syntax鈥.)
Memory freed by DELETE
operations
but still allocated to a specific table can also be
made available for general re-use by performing a
rolling restart of the cluster. See
Section聽15.5.1, 鈥淧erforming a Rolling Restart of the Cluster鈥.
Error Reporting:
A duplicate key error returns the error message
ERROR 23000: Can't write; duplicate key in
table
'tbl_name
'.
Like other MySQL storage engines, the
NDB
storage engine can handle a
maximum of one AUTO_INCREMENT
column per table. However, in the case of a Cluster
table with no explicit primary key, an
AUTO_INCREMENT
column is
automatically defined and used as a
鈥hidden鈥 primary key. For this reason,
you cannot define a table that has an explicit
AUTO_INCREMENT
column unless that
column is also declared using the PRIMARY
KEY
option.
Attempting to create a table with an
AUTO_INCREMENT
column that is not
the table's primary key, and using the
NDB
storage engine, fails with an
error.
Transaction Handling:
NDB Cluster
supports only the
READ COMMITTED
transaction
isolation level.
There is no partial rollback of transactions. A duplicate key or similar error results in a rollback of the entire transaction.
Important: If a
SELECT
from a Cluster table
includes a BLOB
or
TEXT
column, the READ
COMMITTED
transaction isolation level is
converted to a read with read lock. This is done to
guarantee consistency, due to the fact that parts of
the values stored in columns of these types are
actually read from a separate table.
As noted elsewhere in this chapter, MySQL Cluster does not handle large transactions well; it is better to perform a number of small transactions with a few operations each than to attempt a single large transaction containing a great many operations.
Among other considerations, large transactions require very large amounts of memory. Because of this, the transactional behaviour of a number of MySQL statements is effected as described in the following list:
TRUNCATE
is not transactional
when used on NDB
tables. If a
TRUNCATE
fails to empty the
table, then it must be re-run until it is
successful.
DELETE FROM
(even with no
WHERE
clause)
is transactional. For tables
containing a great many rows, you may find that
performance is improved by using several
DELETE FROM ... LIMIT ...
statements to 鈥chunk鈥 the delete
operation. If the objective is to empty the table,
then you may wish to use
TRUNCATE
instead.
LOAD DATA INFILE
is not
transactional when used on NDB
tables. During such an operation, the
NDB
engine can and does commit
at will.
LOAD DATA FROM MASTER
is not
supported in MySQL Cluster.
When copying a table as part of an ALTER
TABLE
, the creation of the copy is
non-transactional. (In any case, this operation is
rolled back when the copy is deleted.)
Node Start, Stop, or Restart:: Starting, stopping, or restarting a node may give rise to temporary errors causing some transactions to fail. These include the following cases:
When first starting a node, it is possible that you may see Error 1204 Temporary failure, distribution changed and similar temporary errors.
The stopping or failure of any data node can result in a number of different node failure errors. (However, there should be no aborted transactions when performing a planned shutdown of the cluster.)
In either of these cases, any errors that are generated must be handled within the application. This should be done by retrying the transaction.
A number of hard limits exist which are configurable, but available main memory in the cluster sets limits. See the complete list of configuration parameters in Section聽15.4.4, 鈥淐onfiguration File鈥. Most configuration parameters can be upgraded online. These hard limits include:
Database memory size and index memory size
(DataMemory
and
IndexMemory
, respectively).
DataMemory
is allocated as 32KB
pages. As each DataMemory
page is
used, it is assigned to a specific table; once
allocated, this memory cannot be freed except by
dropping the table.
See Section聽15.4.4.5, 鈥淒efining Data Nodes鈥,
for further information about
DataMemory
and
IndexMemory
.
The maximum number of operations that can be performed
per transaction is set using the configuration
parameters
MaxNoOfConcurrentOperations
and
MaxNoOfLocalOperations
. Note that
bulk loading, TRUNCATE TABLE
, and
ALTER TABLE
are handled as special
cases by running multiple transactions, and so are not
subject to this limitation.
Different limits related to tables and indexes. For
example, the maximum number of ordered indexes per
table is determined by
MaxNoOfOrderedIndexes
.
Database names, table names and attribute names cannot be
as long in NDB
tables as with other
table handlers. Attribute names are truncated to 31
characters, and if not unique after truncation give rise
to errors. Database names and table names can total a
maximum of 122 characters. (That is, the maximum length
for an NDB Cluster
table name is 122
characters less the number of characters in the name of
the database of which that table is a part.)
The maximum number of tables in a Cluster database is limited to 20320.
In MySQL 5.1.10 and earlier versions, the maximum number
of tables having AUTO_INCREMENT
columns
鈥 including those belonging to hidden primary keys
鈥 is 2048.
This limitation was lifted in MySQL 5.1.11.
The maximum number of attributes per table is limited to 128.
The maximum permitted size of any one row is 8KB. Note
that each BLOB
or
TEXT
column contributes 256 + 8 = 264
bytes towards this total.
The maximum number of attributes per key is 32.
Unsupported features (do not cause errors, but are not supported or enforced):
The foreign key construct is ignored, just as it is in
MyISAM
tables.
Savepoints and rollbacks to savepoints are ignored as in
MyISAM
.
OPTIMIZE
operations are not supported.
LOAD TABLE ... FROM MASTER
is not
supported.
Performance and limitation-related issues:
There are query performance issues due to sequential
access to the NDB
storage engine; it is
also relatively more expensive to do many range scans than
it is with either MyISAM
or
InnoDB
.
The Records in range
statistic is
available but not completely tested or officially
supported. This may result in non-optimal query plans in
some cases. If necessary, you can employ USE
INDEX
or FORCE INDEX
to alter
the execution plan. See Section聽13.2.7.2, 鈥淚ndex Hint Syntax鈥.
Unique hash indexes created with USING
HASH
cannot be used for accessing a table if
NULL
is given as part of the key.
SQL_LOG_BIN
has no effect on data
operations; however, it is supported for schema
operations.
MySQL Cluster cannot produce a binlog for tables having
BLOB
columns but no primary key.
Only the following schema operations are logged in a cluster binlog which is not on the mysqld executing the statement:
CREATE TABLE
ALTER TABLE
DROP TABLE
CREATE DATABASE
/ CREATE
SCHEMA
DROP DATABASE
/ DROP
SCHEMA
CREATE TABLESPACE
ALTER TABLESPACE
DROP TABLESPACE
CREATE LOGFILE GROUP
ALTER LOGFILE GROUP
DROP LOGFILE GROUP
Missing features:
The only supported isolation level is READ
COMMITTED
. (InnoDB supports READ
COMMITTED
, READ UNCOMMITTED
,
REPEATABLE READ
, and
SERIALIZABLE
.) See
Section聽15.8.5, 鈥淏ackup Troubleshooting鈥,
for information on how this can affect backup and restore
of Cluster databases.
No durable commits on disk. Commits are replicated, but there is no guarantee that logs are flushed to disk on commit.
Problems relating to multiple MySQL
servers (not relating to MyISAM
or InnoDB
):
ALTER TABLE
is not fully locking when
running multiple MySQL servers (no distributed table
lock).
DDL operations are not node failure safe. If a node fails
while trying to peform one of these (such as
CREATE TABLE
or ALTER
TABLE
), the data dictionary is locked and no
further DDL statements can be executed without restarting
the cluster.
Issues exclusive to MySQL
Cluster (not related to MyISAM
or InnoDB
):
All machines used in the cluster must have the same architecture. That is, all machines hosting nodes must be either big-endian or little-endian, and you cannot use a mixture of both. For example, you cannot have a management node running on a PowerPC which directs a data node that is running on an x86 machine. This restriction does not apply to machines simply running mysql or other clients that may be accessing the cluster's SQL nodes.
Online adding or dropping of nodes is not possible (the cluster must be restarted in such cases).
While it is possible to run multiple cluster processes concurrently on a single host, it is not always advisable to do so for reasons of performance and high availability, as well as other considerations. In particular, we do not in MySQL 5.1 support for production use any MySQL Cluster deployment in which more than one ndbd process is run on a single physical machine.
We may support multiple data nodes per host in a future MySQL release, following additional testing. However, in MySQL 5.1, such configurations can be considered experimental only.
Issues relating to Disk Data tables:
Disk data objects are subject to the following maximums:
Maxmimum number of tablespaces: 2^32 (4294967296)
Maximum number of data files per tablespace: 2^16 (65535)
Maxmimum data file size: 2^47 (128GB)
Use of Disk Data tables is not supported when running the cluster in diskless mode. Beginning with MySQL 5.1.12, it is disallowed altogether. (Bug#20008)
When using multiple management servers:
You must give nodes explicit IDs in connectstrings because automatic allocation of node IDs does not work across multiple management servers.
You must take extreme care to have the same configurations for all management servers. No special checks for this are performed by the cluster.
Multiple network addresses per data node are not supported. Use of these is liable to cause problems: In the event of a data node failure, an SQL node waits for confirmation that the data node went down but never receives it because another route to that data node remains open. This can effectively make the cluster inoperable.
It is possible to use multiple network hardware interfaces
(such as Ethernet cards) for a single data node, but these
must be bound to the same address. This also means that it
not possible to use more than one [TCP]
section per connection in the
config.ini
file. See
Section聽15.4.4.7, 鈥淐luster TCP/IP Connections鈥, for more
information.
The maximum number of data nodes is 48.
The total maximum number of nodes in a MySQL Cluster is 63. This number includes all SQL nodes (MySQL Servers), API nodes (applications accessing the cluster other than MySQL servers), data nodes, and management servers.
The maximum number of metadata objects in MySQL 5.1 Cluster is 20320. This limit is hard-coded.
MySQL Cluster issues from previous versions that have been resolved in MySQL 5.1:
The NDB Cluster
storage engine now
supports variable-length column types for in-memory
tables.
Previously, this meant that 鈥 for example 鈥
any Cluster table having one or more
VARCHAR
fields which contained only
relatively small values, much more memory and disk space
were required when using the NDBCluster
storage engine than would have been the case for the same
table and data using the MyISAM
engine.
In other words, in the case of a
VARCHAR
column, such a column required
the same amount of storage as a CHAR
column of the same size. In MySQL 5.1, this is no longer
the case for in-memory tables, where storage requirements
for variable-length column types such as
VARCHAR
and BINARY
are comparable to those for these column types when used
in MyISAM
tables (see
Section聽11.5, 鈥淒ata Type Storage Requirements鈥).
Important: For MySQL Cluster Disk Data tables, the fixed-width limitation continues to apply. See Section聽15.11, 鈥淢ySQL Cluster Disk Data Tables鈥.
It is now possible to use MySQL replication with Cluster databases. For details, see Section聽15.10, 鈥淢ySQL Cluster Replication鈥.
However, circular replication is not currently supported for MySQL Cluster. See Section聽15.10.3, 鈥淜nown Issues鈥.
Autodiscovery of databases is now supported for multiple MySQL servers accessing the same MySQL Cluster, provided that a given mysqld is already running and is connected to the cluster at the time that the database is created on a different mysqld.
What this means is that if a mysqld
process first connects to the cluster after a database
named db_name
has been created,
you should issue a CREATE SCHEMA
statement on
the 鈥new鈥 MySQL server when it first accesses
that MySQL Cluster. Once this has been done, the
鈥new鈥 mysqld should be
able to detect any tables in that database tables without
errors.
db_name
This also means that online schema changes in
NDB
tables are now possible. That is,
the result of operations such as ALTER
TABLE
and CREATE INDEX
performed on one SQL node in the cluster are now visible
to the cluster's other SQL nodes without any additional
action being taken.
Beginning with MySQL 5.1.10, it is possible to perform a Cluster backup and restore between different architectures. Previously 鈥 for example 鈥 you could not back up a cluster running on a big-endian platform and then restore from that backup to a cluster running on a little-endian system. (Bug#19255)
Beginning with MySQL 5.1.10, it is possible to install
MySQL with Cluster support to a non-default location and
change the search path for font description files using
either the --basedir
or
--character-sets-dir
options.
(Previously, ndbd in MySQL 5.1 searched
only the default path 鈥 typically
/usr/local/mysql/share/mysql/charsets
鈥 for character sets.)
In MySQL 5.1, it is no longer necessary, when running multiple management servers, to restart all the cluster's data nodes to enable the management nodes to see one another.
In this section, we discuss changes in the implementation of MySQL Cluster in MySQL 5.1 as compared to MySQL 5.0.
There are a number of significant changes in the implementation of
the NDB Cluster
storage engine in MySQL 5.1 as
compared to that in MySQL 5.0. For an overview of these changes,
see Section聽15.14.1, 鈥淢ySQL Cluster Changes in MySQL 5.1鈥
A number of new features for MySQL Cluster have been implemented in MySQL 5.1:
Integration of MySQL Cluster into MySQL replication: This makes it possible to update from any MySQL Server in the cluster and still have the MySQL Replication handled by one of the MySQL Servers in the cluster, with the state of the slave side remaining consistent with the cluster acting as the master.
Support for disk-based records: Records on disk are now supported. Indexed fields including the primary key hash index must still be stored in RAM but all other fields can be on disk.
Variable-sized records: A
column defined as VARCHAR(255)
currently
uses 260 bytes of storage independent of what is stored in
any particular record. In MySQL 5.1 Cluster tables, only the
portion of the column actually taken up by the record will
be stored. This will make possible a reduction in space
requirements for such columns by a factor of 5 in many
cases.
User-defined partitioning:
Users can define partitions based on columns that are part
of the primary key. It is possible to partition
NDB
tables based on
KEY
and LINEAR KEY
schemes. This feature is also available for many other MySQL
storage engines, which support additional partitioning types
that are not available with NDB Cluster
tables.
For additional general information about user-defined partitioning in MySQL 5.1, see Chapter聽16, Partitioning. Specifics of partitioning types are discussed in Section聽16.2, 鈥淧artition Types鈥.
The MySQL Server can also determine whether it is possible
to 鈥prune away鈥 some of the partitions from the
WHERE
clause. See
Section聽16.4, 鈥淧artition Pruning鈥.
Autodiscovery of table schema
changes. In MySQL 5.1, you no longer need to
issue FLUSH TABLES
or a
鈥dummy鈥 SELECT
in order for
new NDB
tables or changes made to schemas
of existing NDB
tables on one SQL node to
be visible on the cluster's other SQL nodes.
Note: When creating a new
database, it is still necessary to issue the CREATE
DATABASE
or CREATE SCHEMA
statement on each SQL node in the cluster.
See MySQL Cluster issues from previous versions that have been resolved in MySQL 5.1 for more information.
The following terms are useful to an understanding of MySQL Cluster or have specialized meanings when used in relation to it.
Cluster:
In its generic sense, a cluster is a set of computers functioning as a unit and working together to accomplish a single task.
NDB
Cluster
:
This is the storage engine used in MySQL to implement data storage, retrieval, and management distributed among several computers.
MySQL Cluster:
This refers to a group of computers working together using the
NDB
storage engine to support a distributed
MySQL database in a shared-nothing
architecture using in-memory
storage.
Configuration files:
Text files containing directives and information regarding the cluster, its hosts, and its nodes. These are read by the cluster's management nodes when the cluster is started. See Section聽15.4.4, 鈥淐onfiguration File鈥, for details.
Backup:
A complete copy of all cluster data, transactions and logs, saved to disk or other long-term storage.
Restore:
Returning the cluster to a previous state, as stored in a backup.
Checkpoint:
Generally speaking, when data is saved to disk, it is said
that a checkpoint has been reached. More specific to Cluster,
it is a point in time where all committed transactions are
stored on disk. With regard to the NDB
storage engine, there are two types of checkpoints which work
together to ensure that a consistent view of the cluster's
data is maintained:
Local Checkpoint (LCP):
This is a checkpoint that is specific to a single node; however, LCP's take place for all nodes in the cluster more or less concurrently. An LCP involves saving all of a node's data to disk, and so usually occurs every few minutes. The precise interval varies, and depends upon the amount of data stored by the node, the level of cluster activity, and other factors.
Global Checkpoint (GCP):
A GCP occurs every few seconds, when transactions for all nodes are synchronized and the redo-log is flushed to disk.
Cluster host:
A computer making up part of a MySQL Cluster. A cluster has both a physical structure and a logical structure. Physically, the cluster consists of a number of computers, known as cluster hosts (or more simply as hosts. See also Node and Node group below.
Node:
This refers to a logical or functional unit of MySQL Cluster, and is sometimes also referred to as a cluster node. In the context of MySQL Cluster, we use the term 鈥node鈥 to indicate a process rather than a physical component of the cluster. There are three node types required to implement a working MySQL Cluster:
Management (MGM) nodes:
Manages the other nodes within the MySQL Cluster. It provides configuration data to the other nodes; starts and stops nodes; handles network partitioning; creates backups and restores from them, and so forth.
SQL (MySQL server) nodes:
Instances of MySQL Server which serve as front ends to data kept in the cluster's data nodes. Clients desiring to store, retrieve, or update data can access an SQL node just as they would any other MySQL Server, employing the usual authentication methods and API's; the underlying distribution of data between node groups is transparent to users and applications. SQL nodes access the cluster's databases as a whole without regard to the data's distribution across different data nodes or cluster hosts.
Data nodes:
These nodes store the actual data. Table data fragments are stored in a set of node groups; each node group stores a different subset of the table data. Each of the nodes making up a node group stores a replica of the fragment for which that node group is responsible. Currently, a single cluster can support up to 48 data nodes total.
It is possible for more than one node to co-exist on a single machine. (In fact, it is even possible to set up a complete cluster on one machine, although one would almost certainly not want to do this in a production environment.) It may be helpful to remember that, when working with MySQL Cluster, the term host refers to a physical component of the cluster whereas a node is a logical or functional component (that is, a process).
Note Regarding Terms: In older versions of the MySQL Cluster documentation, data nodes were sometimes referred to as 鈥database nodes鈥. The term 鈥storage nodes鈥 has also been used. In addition, SQL nodes were sometimes known as 鈥client nodes鈥. This older terminology has been deprecated to minimize confusion, and for this reason should be avoided. They are also often referred to as 鈥API nodes鈥 鈥 an SQL node is actually an API node that provides an SQL interface to the cluster.
Node group:
A set of data nodes. All data nodes in a node group contain the same data (fragments), and all nodes in a single group should reside on different hosts. It is possible to control which nodes belong to which node groups.
For more information, see Section聽15.2.1, 鈥淢ySQL Cluster Nodes, Node Groups, Replicas, and Partitions鈥.
Node failure:
MySQL Cluster is not solely dependent upon the functioning of any single node making up the cluster; the cluster can continue to run if one or more nodes fail. The precise number of node failures that a given cluster can tolerate depends upon the number of nodes and the cluster's configuration.
Node restart:
The process of restarting a failed cluster node.
Initial node restart:
The process of starting a cluster node with its filesystem removed. This is sometimes used in the course of software upgrades and in other special circumstances.
System crash (or system failure):
This can occur when so many cluster nodes have failed that the cluster's state can no longer be guaranteed.
System restart:
The process of restarting the cluster and reinitializing its state from disk logs and checkpoints. This is required after either a planned or an unplanned shutdown of the cluster.
Fragment:
A portion of a database table; in the NDB
storage engine, a table is broken up into and stored as a
number of fragments. A fragment is sometimes also called a
鈥partition鈥; however, 鈥fragment鈥 is
the preferred term. Tables are fragmented in MySQL Cluster in
order to facilitate load balancing between machines and nodes.
Replica:
Under the NDB
storage engine, each table
fragment has number of replicas stored on other data nodes in
order to provide redundancy. Currently, there may be up four
replicas per fragment.
Transporter:
A protocol providing data transfer between nodes. MySQL Cluster currently supports four different types of transporter connections:
TCP/IP
This is, of course, the familiar network protocol that underlies HTTP, FTP (and so on) on the Internet. TCP/IP can be used for both local and remote connections.
SCI
Scalable Coherent Interface is a high-speed protocol used in building multiprocessor systems and parallel-processing applications. Use of SCI with MySQL Cluster requires specialized hardware, as discussed in Section聽15.12.1, 鈥淐onfiguring MySQL Cluster to use SCI Sockets鈥. For a basic introduction to SCI, see this essay at dolphinics.com.
SHM
Unix-style shared
memory segments. Where
supported, SHM is used automatically to connect nodes
running on the same host. The
Unix
man page for shmop(2)
is a good
place to begin obtaining additional information about this
topic.
Note: The cluster transporter is internal to the cluster. Applications using MySQL Cluster communicate with SQL nodes just as they do with any other version of MySQL Server (via TCP/IP, or through the use of Unix socket files or Windows named pipes). Queries can be sent and results retrieved using the standard MySQL client APIs.
NDB
:
This stands for Network
Database,
and refers to the storage engine used to enable MySQL Cluster.
The NDB
storage engine supports all the
usual MySQL data types and SQL statements, and is
ACID-compliant. This engine also provides full support for
transactions (commits and rollbacks).
shared-nothing architecture:
The ideal architecture for a MySQL Cluster. In a true shared-nothing setup, each node runs on a separate host. The advantage such an arrangement is that there no single host or node can act as single point of failure or as a performance bottle neck for the system as a whole.
In-memory storage:
All data stored in each data node is kept in memory on the node's host computer. For each data node in the cluster, you must have available an amount of RAM equal to the size of the database times the number of replicas, divided by the number of data nodes. Thus, if the database takes up 1GB of memory, and you want to set up the cluster with four replicas and eight data nodes, a minimum of 500MB memory will be required per node. Note that this is in addition to any requirements for the operating system and any other applications that might be running on the host.
In MySQL 5.1, it is also possible to create Disk Data tables where non-indexed columns are stored on disk, thus reducing the memory footprint required by the cluster. Note that indexes and indexed column data are still stored in RAM. See Section聽15.11, 鈥淢ySQL Cluster Disk Data Tables鈥.
Table:
As is usual in the context of a relational database, the term 鈥table鈥 denotes a set of identically structured records. In MySQL Cluster, a database table is stored in a data node as a set of fragments, each of which is replicated on additional data nodes. The set of data nodes replicating the same fragment or set of fragments is referred to as a node group.
Cluster programs:
These are command-line programs used in running, configuring, and administering MySQL Cluster. They include both server daemons:
ndbd:
The data node daemon (runs a data node process)
ndb_mgmd:
The management server daemon (runs a management server process)
and client programs:
ndb_mgm:
The management client (provides an interface for executing management commands)
ndb_waiter:
Used to verify status of all nodes in a cluster
ndb_restore:
Restores cluster data from backup
For more about these programs and their uses, see Section聽15.6, 鈥淧rocess Management in MySQL Cluster鈥.
Event log:
MySQL Cluster logs events by category (startup, shutdown, errors, checkpoints, and so on), priority, and severity. A complete listing of all reportable events may be found in Section聽15.7.3, 鈥淓vent Reports Generated in MySQL Cluster鈥. Event logs are of two types:
Cluster log:
Keeps a record of all desired reportable events for the cluster as a whole.
Node log:
A separate log which is also kept for each individual node.
Under normal circumstances, it is necessary and sufficient to keep and examine only the cluster log. The node logs need be consulted only for application development and debugging purposes.