Chapter 15. MySQL Cluster

Table of Contents

15.1. MySQL Cluster Overview
15.2. Basic MySQL Cluster Concepts
15.2.1. MySQL Cluster Nodes, Node Groups, Replicas, and Partitions
15.3. Simple Multi-Computer How-To
15.3.1. Hardware, Software, and Networking
15.3.2. Multi-Computer Installation
15.3.3. Multi-Computer Configuration
15.3.4. Initial Startup
15.3.5. Loading Sample Data and Performing Queries
15.3.6. Safe Shutdown and Restart
15.4. MySQL Cluster Configuration
15.4.1. Building MySQL Cluster from Source Code
15.4.2. Installing the Software
15.4.3. Quick Test Setup of MySQL Cluster
15.4.4. Configuration File
15.5. Process Management in MySQL Cluster
15.5.1. MySQL Server Process Usage for MySQL Cluster
15.5.2. ndbd, the Storage Engine Node Process
15.5.3. ndb_mgmd, the Management Server Process
15.5.4. ndb_mgm, the Management Client Process
15.5.5. Command Options for MySQL Cluster Processes
15.6. Management of MySQL Cluster
15.6.1. MySQL Cluster Startup Phases
15.6.2. Commands in the Management Client
15.6.3. Event Reports Generated in MySQL Cluster
15.6.4. Single-User Mode
15.6.5. On-line Backup of MySQL Cluster
15.7. Using High-Speed Interconnects with MySQL Cluster
15.7.1. Configuring MySQL Cluster to use SCI Sockets
15.7.2. Understanding the Impact of Cluster Interconnects
15.8. Known Limitations of MySQL Cluster
15.9. MySQL Cluster Development Roadmap
15.9.1. MySQL Cluster Changes in MySQL 5.0
15.9.2. MySQL 5.1 Development Roadmap for MySQL Cluster
15.10. MySQL Cluster FAQ
15.11. MySQL Cluster Glossary

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 5.0 binary releases and in RPMs compatible with most modern Linux distributions. (If you install using RPM files, note that both the mysql-server and mysql-max RPMs must be installed to have MySQL Cluster capability.)

The operating systems on which MySQL Cluster is currently available are Linux, Mac OS X, and Solaris. (Some users have reported success with running MySQL Cluster on FreeBSD, although this is not yet officially supported by MySQL AB.) We are working to make Cluster run on all operating systems supported by MySQL, including Windows, and will update this page as new platforms are supported.

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

Additional resources

15.1. MySQL Cluster Overview

MySQL Cluster is a technology that enables clustering of in-memory databases in a share-nothing system. The share-nothing architecture allows the system to work with very inexpensive hardware, and without any specific requirements on hardware or software. It also does not have any single point of failure because each component has its own memory and disk.

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 “MySQL Cluster” refers to the combination of MySQL and the NDB storage engine.

A MySQL Cluster consists of a set of computers, each running a number of processes including MySQL servers, data nodes for NDB Cluster, management servers, and (possibly) specialized data access programs. The relationship of these components in a cluster is shown here:

MySQL Cluster Components

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.

By bringing MySQL Cluster to the Open Source world, MySQL AB makes clustered data management with high availability, high performance, and scalability available to all who need it.

15.2. Basic MySQL Cluster Concepts

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.

We will now describe how to set up a MySQL Cluster consisting of an NDB storage engine and some MySQL servers.

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 “node” is used to indicate a computer, but when discussing MySQL Cluster it means a process. There can be any number of nodes on a single computer, for which we use the term cluster host.

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:

  • The management node (MGM node): The role of this type of node is to manage the other nodes within the MySQL Cluster, such 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.

  • The data node: This is the type of node that stores the cluster's 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.

  • The SQL node: This is the node that accesses the cluster data. In the case of MySQL Cluster, a client 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.

For a brief introduction to the relationships between nodes, node groups, replicas, and partitions in MySQL Cluster, see Section 15.2.1, “MySQL Cluster Nodes, Node Groups, Replicas, and Partitions”.

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 storage 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.

15.2.1. MySQL Cluster Nodes, Node Groups, Replicas, and Partitions

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 —that is, a copy of the partition (see below) assigned to the node group of which the node is a member.

    Each data node is usually located on a separate computer. However, it is also possible to host multiple data nodes on a single computer having more than one processor. In such cases, it is feasible to run one instance of ndbd per physical CPU. (Note that a processor with multiple cores is still a single processor.)

    It is common for the terms “node” and “data node” to be used interchangeably when referring to an ndbd process; where mentioned, management 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 a partition, or set 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 node groups participating in the cluster, and each node group is responible for keeping at least one copy of the partition assigned to it (that is, at least one replica) available to the cluster.

  • 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 following diagram illustrates a MySQL Cluster with four data nodes, arranged in two node groups of two nodes each. Note that no nodes other than data 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.

A MySQL Cluster, with 2 node groups having 2
          nodes each

The data stored by the cluster is divided into two partitions, labeled A and B in the diagram. Each partition is stored — in multiple copies — on a node group. The data making up Partition A is stored on Node A-1, and this data is identical to that stored by Node A-2. The data stored by Nodes B-1 and B-2 is also the same — these two nodes store identical copies of the data making up Partition B.

What this means so far as the continued operation of a MySQL Cluster is this: so long as each node group participating in the cluster has at least one “live” node, the cluster has a complete copy of all data and remains viable. This is illustrated in the next diagram.

Nodes required to keep a 2x2 cluster

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 A and at least one node in Node Group B is sufficient to keep the cluster “alive” (indicated by arrows in the diagram). However, if both nodes from either node group fail, the remaining two nodes are not sufficient (shown by 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.

15.3. Simple Multi-Computer How-To

This section is a “How-To” that describes the basics for how to plan, install, configure, and run a MySQL Cluster. Whereas the examples in Section 15.4, “MySQL Cluster Configuration” 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:

  1. 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:

    NodeIP Address
    Management (MGM) node192.168.0.10
    MySQL server (SQL) node192.168.0.20
    Data (NDBD) node "A"
    Data (NDBD) node "B"

    This may be made clearer in the following diagram:

    MySQL Cluster Multi-Computer

    Note: 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.

  2. 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.

  3. 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.

  4. 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 either Solaris or Mac OS X. 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 15.3.1, “Hardware, Software, and Networking”.)

15.3.1. Hardware, Software, and Networking

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 subject to change, and that we intend to implement disk-based storage in a future MySQL Cluster release.) Naturally, multiple and faster CPUs will enhance performance. Memory requirements for 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 Mac OS X or Solaris, the standard installation is sufficient. For Linux, a standard, “out of the box” installation should be all that is necessary. The MySQL software requirements are simple: all that is needed is a production release of MySQL-max 5.0; you must use the -max version of MySQL to have Cluster support. (See Section 5.3, “The mysqld-max Extended MySQL Server”.) 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 -max binary appropriate to your Linux, Solaris, or Mac OS X operating system, available via the MySQL software downloads page at

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 15.7, “Using High-Speed Interconnects with MySQL Cluster”, for more about this protocol and its use with MySQL Cluster.

15.3.2. Multi-Computer Installation

Each MySQL Cluster host computer running storage or SQL nodes must have installed on it a MySQL-max binary. For management nodes, it is not necessary to install the MySQL server binary, but you do have to install the MGM server daemon and client binaries (ndb_mgmd and ndb_mgm, respectively). 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-max-5.0.19-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 2.8.3, “Installing from the Development Source Tree”.)

RPMs are also available for both 32-bit and 64-bit Linux platforms; as of MySQL 4.1.10a, the -max binaries installed by the RPMs support the NDBCluster storage engine. If you choose to use these rather than the binary files, be aware that you must install both the -server and -max packages on all machines that are to host cluster nodes. (See Section 2.4, “Installing MySQL on Linux”, for more information about installing MySQL using the RPMs.) After installing from RPM, you will still need to configure the cluster as discussed in Section 15.3.3, “Multi-Computer Configuration”.

Note: After completing the installation, do not yet start any of the binaries. We will show you how to do so following the configuration of all nodes.

Storage and SQL Node Installation

On each of the three machines designated to host storage or SQL nodes, perform the following steps as the system root user:

  1. Check your /etc/passwd and /etc/group files (or use whatever tools are provided by your operating system for manging 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.

  2. Change location to the directory containing the downloaded file, unpack the archive, and create a symlink to the mysql-max directory named mysql. Note that the actual file and directory names will vary according to the MySQL version number.

    shell> cd /var/tmp
    shell> tar -xzvf -C /usr/local mysql-max-5.0.19-pc-linux-gnu-i686.tar.gz
    shell> ln -s /usr/local/mysql-max-5.0.19-pc-linux-gnu-i686 /usr/local/mysql
  3. 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
  4. 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 15.3.3, “Multi-Computer Configuration”.)

  5. 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 — 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 for each machine on which a storage or SQL node is to reside.

Management Node Installation

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 -max 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:

  1. 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:

    shell> cd /var/tmp
    shell> tar -zxvf mysql-max-5.0.19-pc-linux-gnu-i686.tar.gz \
              /usr/local '*/bin/ndb_mgm*'
  2. Change location to the directory into which you unpacked the files, and then make both of them executable:

    shell> cd /usr/local
    shell> chmod +x ndb_mgm*

In Section 15.3.3, “Multi-Computer Configuration”, we will create and write configuration files for all of the nodes in our example Cluster.

15.3.3. Multi-Computer Configuration

For our four-node, four-host MySQL Cluster, we will need 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 telling 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, “The MySQL Cluster connectstring.

  • 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:
ndbcluster                      # run NDB engine
ndb-connectstring=  # location of MGM node

# Options for ndbd process:
ndb-connectstring=  # location of MGM node

After entering the preceding information, save this file and exit the text editor. Do this for the machines hosting data node “A”, data node “B”, and the SQL node.

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:
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:
portnumber=2202   # This the default; however, you can use any
                  # port that is free for all the hosts in cluster
                  # Note: It is recommended beginning with MySQL 5.0 that
                  # you do not specify the portnumber at all and simply allow
                  # the default value to be used instead

# Management process options:
hostname=           # Hostname or IP address of MGM node
datadir=/var/lib/mysql-cluster  # Directory for MGM node logfiles

# Options for data node "A":
                                # (one [NDBD] section per data node)
hostname=           # Hostname or IP address
datadir=/usr/local/mysql/data   # Directory for this data node's datafiles

# Options for data node "B":
hostname=           # Hostname or IP address
datadir=/usr/local/mysql/data   # Directory for this data node's datafiles

# SQL node options:
hostname=           # 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, where it can be found listed under “Examples.”)

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 15.3.4, “Initial Startup”.

For more detailed information about the available MySQL Cluster configuration parameters and their uses, see Section 15.4.4, “Configuration File”, and Section 15.4, “MySQL Cluster Configuration”. For configuration of MySQL Cluster as relates to making backups, see Section, “Configuration for Cluster Backup”.

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.

15.3.4. Initial Startup

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. Although it is possible to start the nodes in any order, it is recommended that the management node be started first, followed by the storage nodes, and then finally by any SQL nodes:

  1. On the management host, issue the following command from the system shell to start the MGM 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 15.5.3, “ndb_mgmd, the Management Server Process”, for details.)

  2. 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.

  3. 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 startup script installed in /etc/init.d to start the MySQL server process on the SQL node. Note that you need to install the -max server RPM in addition to the Standard server RPM to run the -max server binary.

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    @  (Version: 5.0.19, Nodegroup: 0, Master)
id=3    @  (Version: 5.0.19, Nodegroup: 0)

[ndb_mgmd(MGM)] 1 node(s)
id=1    @  (Version: 5.0.19)

[mysqld(SQL)]   1 node(s)
id=4   (Version: 5.0.19)

Note: If you are using an older version of MySQL, you may see the SQL node referenced as [mysqld(API)]. This reflects an older usage that is now deprecated.

You should now be ready to work with databases, tables, and data in MySQL Cluster. See Section 15.3.5, “Loading Sample Data and Performing Queries”, for a brief discussion.

15.3.5. Loading Sample Data and Performing Queries

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:


    Alternatively, for an existing table that uses a different storage engine, use ALTER TABLE to change the table to use NDB Cluster:

  • 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):

  `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',

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 will 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:

  `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',

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=engine_name or ENGINE=engine_name with 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

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 5.0 does not support autodiscovery of databases. (See Section 15.8, “Known Limitations of MySQL Cluster”.) 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 (beginning with MySQL 5.0.2, you may use CREATE SCHEMA world instead), followed by FLUSH TABLES on each SQL node in the cluster. This will cause the node to recognize the database and read its table definitions.

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: 5.0.19

Type 'help;' or '\h' for help. Type '\c' to clear the buffer.


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 2.9.3, “Securing the Initial MySQL Accounts”.

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;

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 |
| São Paulo |    9968485 |
| Shanghai  |    9696300 |
| Jakarta   |    9604900 |
5 rows in set (0.34 sec)

mysql> \q


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 MGM or storage nodes. This brief example shows how we might execute the SELECT statement just shown by using PHP 5's mysqli extension running on a Web server elsewhere on the network:

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
  <meta http-equiv="Content-Type"
        content="text/html; charset=iso-8859-1">
  <title>SIMPLE mysqli SELECT</title>
  # connect to SQL node:
  $link = new mysqli('', '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">
    <th width="10%">City</th>
    # 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);
  # ...and verify the number of rows that were retrieved
    printf("<p>Affected rows: %d</p>\n", $link->affected_rows);
    # otherwise, tell us what went wrong
    echo mysqli_error();

  # free the result set and the mysqli connection object

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.

15.3.6. Safe Shutdown and Restart

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 4.3.1, “Using Options on the Command Line”. 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 ( in our example setup):

    shell> ndb_mgmd -f /var/lib/mysql-cluster/config.ini
  • On each of the data node hosts ( and

    shell> ndbd

    Remember not to invoke this command with the --initial option when restarting an NDBD node normally.

  • On the SQL host (

    shell> mysqld &

For information on making Cluster backups, see Section, “Using The Management Server to Create a Backup”.

To restore the cluster from backup requires the use of the ndb_restore command. This is covered in Section, “How to Restore a Cluster Backup”.

More information on configuring MySQL Cluster can be found in Section 15.4, “MySQL Cluster Configuration”.

15.4. MySQL Cluster Configuration

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.

15.4.1. Building MySQL Cluster from Source Code

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 5.0 BitKeeper tree, be sure to use the --with-ndbcluster option when running configure. You can also use the BUILD/compile-pentium-max build script. Note that 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 2.8.3, “Installing from the Development Source Tree”.

15.4.2. Installing the Software

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 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.

15.4.3. Quick Test Setup of MySQL Cluster

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.
NoOfReplicas= 1



DataDir= /var/lib/mysql-cluster


You can now start the ndb_mgmd management server. By default, it atttempts 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 DB node by running ndbd. When starting ndbd for a given DB 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:


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: 5.0.19-Max

Type 'help;' or '\h' for help. Type '\c' to clear the buffer.

*************************** 12. row ***************************
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

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:

Cluster Configuration
[ndbd(NDB)]     1 node(s)
id=2    @  (Version: 3.5.3, Nodegroup: 0, Master)

[ndb_mgmd(MGM)] 1 node(s)
id=1    @  (Version: 3.5.3)

[mysqld(API)]   3 node(s)
id=3    @  (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.

15.4.4. Configuration File

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.

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 1.8, “How to Report Bugs or Problems”. Example Configuration for a MySQL Cluster

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 5.0)

# enable ndbcluster storage engine, and provide connectstring for
# management server host (default port is 1186)

# provide connectstring for management server host (default port: 1186)

# provide connectstring for management server host (default port: 1186)

# provide location of cluster configuration file

(For more information on connectstrings, see Section, “The MySQL Cluster connectstring.)

# 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

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

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=path_name on the ndb_mgmd command line. If the configuration file is not specified, ndb_mgmd by default tries to read a file named 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 (‘:’) as well as the equals sign (‘=’). 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

NoOfReplicas= 2
DataDir= /var/lib/mysql-cluster

DataDir= /var/lib/mysql-cluster




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.

There are six different sections in that you can use in the config.ini configuration file:

  • [COMPUTER]: Defines the cluster hosts.

  • [NDBD]: Defines the cluster's data nodes.

  • [MYSQLD]: Defines the cluster's MySQL server nodes.

  • [MGM] or [NDB_MGMD]: Defines the cluster's management server node.

  • [TCP]: Defines TCP/IP connections between nodes in the cluster, with TCP/IP being the default connection protocol.

  • [SHM]: Defines shared-memory connections between nodes. Formerly, this type of connection was available only in binaries that were built using the --with-ndb-shm option. In MySQL 5.0-Max, it is enabled by default, but should still be considered experimental.

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. The MySQL Cluster connectstring

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> := 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,"
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 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:

    1. Set the NDB_CONNECTSTRING environment variable to contain the connectstring.

    2. 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. Defining the Computers Making up a MySQL Cluster

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.

  • Id

    This is an integer value, used to refer to the host computer elsewhere in the configuration file.

  • HostName

    This is the computer's hostname or IP address. Defining the MySQL Cluster Management Server

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.

  • Id

    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.

  • ExecuteOnComputer

    This refers to one of the computers defined in the [COMPUTER] section.

  • PortNumber

    This is the port number on which the management server listens for configuration requests and management commands.

  • LogDestination

    This parameter specifies where to send cluster logging information. There are three options in this regard: CONSOLE, SYSLOG, and FILE:

    • CONSOLE outputs the log to stdout:

    • 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.

    • 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 logfile.

      • 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 logfile 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 logfiles.


      It is possible to specify multiple log destinations separated by semicolons as shown here:


      The default value for the FILE parameter is FILE:filename=ndb_node_id_cluster.log,maxsize=1000000,maxfiles=6, where node_id is the ID of the node.

  • ArbitrationRank

    This parameter is used to define which nodes can act as arbitrators. Only MGM 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.

  • ArbitrationDelay

    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.

  • DataDir

    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.) Defining MySQL Cluster Data Nodes

The [NDBD] section is 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.

  • The parameter NoOfReplicas

These mandatory parameters must be defined in the [NDBD DEFAULT] section.

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. HostName, Id and ExecuteOnComputer must be defined in the local [NDBD] section.

Identifying Data Nodes

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.

For each parameter it is possible to use K, M, or G as a suffix to indicate units of 1024, 1024×1024, or 1024×1024×1024. (For example, 100K means 100 × 1024 = 102400.) Parameter names and values are currently case-sensitive.

  • Id

    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.

  • ExecuteOnComputer

    This refers to one of the computers (hosts) defined in the COMPUTER section.

  • HostName

    Specifying this parameter has an effect similar to specifying ExecuteOnComputer. It defines the hostname of the computer on which the storage node is to reside. To specify a hostname other than localhost, either this parameter or ExecuteOnComputer is required.

  • ServerPort (OBSOLETE)

    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.

  • NoOfReplicas

    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.

  • DataDir

    This parameter specifies the directory where trace files, log files, pid files and error logs are placed.

  • FileSystemPath

    This parameter specifies the directory where all files created for metadata, REDO logs, UNDO logs 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.

  • BackupDataDir

    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 and Index 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:

  • DataMemory

    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. Each record is currently of fixed size. (Even VARCHAR columns are stored as fixed-width columns.) 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. 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. 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 2–4GB 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. In addition, it is also possible to run more than one ndbd process per machine, and this may prove advantageous on machines with multiple CPUs.

  • IndexMemory

    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.

The following example illustrates how memory is used for a table. Consider this table definition:

CREATE TABLE example (

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 × 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.

Transaction Parameters

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.

  • MaxNoOfConcurrentTransactions

    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.

  • MaxNoOfConcurrentOperations

    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.

  • MaxNoOfLocalOperations

    By default, this parameter is calculated as 1.1 × 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.

Transaction Temporary Storage

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 parallelism is not required for the cluster.

  • MaxNoOfFiredTriggers

    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.

  • TransactionBufferMemory

    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.

    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 × 128 bytes (500KB). A similar buffer for key information, ZDATABUF_FILESIZE (also in Dbtc.hpp) contains 4000 × 16 = 62.5KB of buffer space. Dbtc is the module that handles transaction coordination.

Scans and Buffering

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 × 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.

The default value for TransactionBufferMemory is 1MB.

  • MaxNoOfConcurrentScans

    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.

    This parameter specifies the number of scans possible in the transaction coordinator. 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.

  • MaxNoOfLocalScans

    Specifies the number of local scan records if many scans are not fully parallelized.

  • BatchSizePerLocalScan

    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.

  • LongMessageBuffer

    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.

Logging and Checkpointing

These [NDBD] paramaters control log and checkpoint behavior.

  • NoOfFragmentLogFiles

    This parameter sets the size of the node's REDO log files. REDO log files are organized in a ring. It is extremely important that the first and last log files (sometimes referred to as the “head” and “tail” 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.

    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.

  • MaxNoOfSavedMessages

    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.

Metadata Objects

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 “suggestions” to the cluster, and any that are not specified revert to the default values shown.

  • MaxNoOfAttributes

    Defines the number of attributes that can be defined in the cluster.

    The default value is 1000, with the minimum possible value being 32. There is no maximum. 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 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.

  • MaxNoOfTables

    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.

  • MaxNoOfOrderedIndexes

    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.

  • MaxNoOfUniqueHashIndexes

    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.

  • MaxNoOfTriggers

    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.

    Note: When replication between clusters is supported, this will also make use of internal triggers.

    This parameter sets the maximum number of trigger objects in the cluster.

    The default value is 768.

  • MaxNoOfIndexes

    This parameter is deprecated in MySQL 5.0; 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.

Boolean Parameters

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.

  • LockPagesInMainMemory

    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.

    This feature is disabled by default.

  • StopOnError

    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.

  • Diskless

    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, backups are performed but backup data is not actually stored.

    Diskless is disabled by default.

  • RestartOnErrorInsert

    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.

  • TimeBetweenWatchDogCheck

    To prevent the main thread from getting stuck in an endless loop at some point, a “watchdog” 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).

  • StartPartialTimeout

    This parameter specifies how long the Cluster waits for all storage 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 other words, the cluster may start only if all nodes are available.

  • StartPartitionedTimeout

    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).

  • StartFailureTimeout

    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 means that no data node timeout is applied.

    The default value is 60000 milliseconds (60 seconds). For data nodes containing extremely large amounts of data, this parameter should be increased. For example, in the case of a storage node containing several gigabytes of data, a period as long as 10–15 minutes (that is, 600,000 to 1,000,000 milliseconds) might be required to to perform a node restart.

  • HeartbeatIntervalDbDb

    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.

  • HeartbeatIntervalDbApi

    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 “dead,” 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 storage node watches the MySQL servers connected to it, independently of all other data nodes.

  • TimeBetweenLocalCheckpoints

    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 × 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.

  • TimeBetweenGlobalCheckpoints

    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 — a complete crash of the cluster — 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).

  • TransactionInactiveTimeout

    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 “dead

    • 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).

  • 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 “page” 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. This value includes the writing of UNDO log records for data pages. That is, this parameter handles the limitation of writes from data memory. UNDO log records for index pages are handled by the parameter NoOfDiskPagesToDiskAfterRestartACC. (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.

    The default value is 40 (3.2MB of data pages per second).

  • 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).

  • NoOfDiskPagesToDiskDuringRestartTUP

    This parameter in a similar fashion as 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).

  • 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).

  • ArbitrationTimeout

    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).

Buffering and Logging

Several [NDBD] configuration parameters corresponding to former compile-time parameters are also available. These enable the advanced user to have more control over the resources used by node processes and to adjust various buffer sizes at need.

These buffers are used as front ends to the file system when writing log records to disk. 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 “faked” by the NDB storage engine's filesystem abstraction layer.

  • UndoIndexBuffer

    The UNDO index buffer, whose size is set by this parameter, is used during local checkpoints. The NDB storage engine uses a recovery scheme based on checkpoint consistency in conjunction with an operational REDO log. To produce a consistent checkpoint without blocking the entire system for writes, UNDO logging is done while performing the local checkpoint. UNDO logging is activated on a single table fragment at a time. This optimization is possible because tables are stored entirely in main memory.

    The UNDO index buffer is used for the updates on the primary key hash index. Inserts and deletes rearrange the hash index; the NDB storage engine writes UNDO log records that map all physical changes to an index page so that they can be undone at system restart. It also logs all active insert operations for each fragment at the start of a local checkpoint.

    Reads and updates set lock bits and update a header in the hash index entry. These changes are handled by the page-writing algorithm to ensure that these operations need no UNDO logging.

    This buffer is 2MB by default. The minimum value is 1MB, which is sufficient for most applications. For applications doing extremely large or numerous inserts and deletes together with large transactions and large primary keys, it may be necessary to increase the size of this buffer. If this buffer is too small, the NDB storage engine issues internal error code 677 (Index UNDO buffers overloaded).

  • UndoDataBuffer

    This parameter sets the size of the UNDO data buffer, which performs a function similar to that of the UNDO index buffer, except the UNDO data buffer is used with regard to data memory rather than index memory. This buffer is used during the local checkpoint phase of a fragment for inserts, deletes, and updates.

    Because UNDO log entries tend to grow larger as more operations are logged, this buffer is also larger than its index memory counterpart, with a default value of 16MB.

    This amount of memory may be unnecessarily large for some applications. In such cases, it is possible to decrease this size to a minimum of 1MB.

    It is rarely necessary to increase the size of this buffer. If there is such a need, it is a good idea to check whether the disks can actually handle the load caused by database update activity. A lack of sufficient disk space cannot be overcome by increasing the size of this buffer.

    If this buffer is too small and gets congested, the NDB storage engine issues internal error code 891 (Data UNDO buffers overloaded).

  • RedoBuffer

    All update activities also 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 “fuzzy” checkpoint of the data together with the UNDO log, 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. The minimum value is 1MB.

    If this buffer is too small, the NDB storage engine issues error code 1221 (REDO log buffers overloaded).

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.

  • LogLevelStartup

    The reporting level for events generated during startup of the process.

    The default level is 1.

  • LogLevelShutdown

    The reporting level for events generated as part of graceful shutdown of a node.

    The default level is 0.

  • LogLevelStatistic

    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.

  • LogLevelCheckpoint

    The reporting level for events generated by local and global checkpoints.

    The default level is 0.

  • LogLevelNodeRestart

    The reporting level for events generated during node restart.

    The default level is 0.

  • LogLevelConnection

    The reporting level for events generated by connections between cluster nodes.

    The default level is 0.

  • LogLevelError

    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.

  • LogLevelInfo

    The reporting level for events generated for information about the general state of the cluster.

    The default level is 0.

Backup Parameters

The [NDBD] parameters discussed in this section define memory buffers set aside for execution of online backups.

  • BackupDataBufferSize

    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.

  • BackupLogBufferSize

    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, “Configuration for Cluster Backup”.

    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.

  • BackupMemory

    This parameter is simply the sum of BackupDataBufferSize and BackupLogBufferSize.

    The default value is 2MB + 2MB = 4MB.

  • BackupWriteSize

    This parameter specifies the size of messages written to disk by the backup log and backup data buffers.

    The default value is 32KB. Defining the SQL Nodes in a MySQL Cluster

The [MYSQLD] sections in the config.ini file define the behavior of the MySQL servers (SQL 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 node.

  • Id

    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.

  • ExecuteOnComputer

    This refers to one of the computers (hosts) defined in a [COMPUTER] section of the configuration file.

  • ArbitrationRank

    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.

  • ArbitrationDelay

    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.

  • BatchByteSize

    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.

  • BatchSize

    This parameter is measured in number of records and is by default set to 64. The maximum size is 992.

  • MaxScanBatchSize

    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. MySQL Cluster TCP/IP Connections

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, “MySQL Cluster TCP/IP Connections Using Direct Connections”.) [TCP] sections in the config.ini file explicitly define TCP/IP connections between nodes in the cluster.

It is only necessary to define a connection to override the default connection parameters. In that case, it is necessary to define at least NodeId1, NodeId2, and the parameters to change.

It is also possible to change the default values for these parameters by setting them in the [TCP DEFAULT] section.

  • NodeId1, NodeId2

    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, “Defining the SQL Nodes in a MySQL Cluster”.

  • SendBufferMemory

    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.

  • SendSignalId

    To be able to retrace a distributed message diagram 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.

  • Checksum

    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.

  • PortNumber (OBSOLETE)

    This formerly specified the port number to be used for listening for connections from other nodes. This parameter should no longer be used.

  • ReceiveBufferMemory

    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. MySQL Cluster TCP/IP Connections Using Direct Connections

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

# SQL Node

# Data Nodes


# 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 Shared-Memory Connections

MySQL Cluster attempts to use the shared memory transporter and configure it automatically where possible, chiefly where more than one node runs concurrently on the same cluster host. (In very early versions of MySQL Cluster, shared memory segments functioned only when the -max binary was built using --with-ndb-shm.) [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 5.0. 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.

  • NodeId1, NodeId2

    To identify a connection between two nodes it is necessary to provide node identifiers for each of them, as NodeId1 and NodeId2.

  • ShmKey

    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.

  • ShmSize

    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.

  • SendSignalId

    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.

  • Checksum

    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. MySQL Cluster SCI Transport Connections

[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-Max binaries are built using --with-ndb-sci=/your/path/to/SCI. The path should point to a directory that contains at a minimum lib and include directories containing SISCI libraries and header files. (See Section 15.7, “Using High-Speed Interconnects with MySQL Cluster” 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.

  • NodeId1, NodeId2

    To identify a connection between two nodes it is necessary to provide node identifiers for each of them, as NodeId1 and NodeId2.

  • Host1SciId0

    This identifies the SCI node ID on the first Cluster node (identified by NodeId1).

  • Host1SciId1

    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.

  • Host2SciId0

    This identifies the SCI node ID on the second Cluster node (identified by NodeId2).

  • Host2SciId1

    When using two SCI cards to provide failover, this parameter identifies the second SCI card to be used on the second node.

  • SharedBufferSize

    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.

  • SendLimit

    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.

  • SendSignalId

    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.

  • Checksum

    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.

15.5. Process Management in MySQL Cluster

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:

15.5.1. MySQL Server Process Usage for MySQL Cluster

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 -max binaries available from 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, is the host where the management server resides, and the management server listens for cluster messages on port 1186:

shell> mysqld

See Section, “The MySQL Cluster connectstring, for more information on connectstrings.

Given this information, the MySQL server will be a full participant in the cluster. (We sometimes 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.

15.5.2. ndbd, the Storage Engine Node Process

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 storage 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 storage node whose node ID is 2.

  • ndb_node_id_error.log 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:

    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

    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 file (see below). Error log entries are thus overwritten in a cyclical and not sequential fashion.

  • ndb_node_id_trace.log.trace_id 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.

    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.

  • is the file that keeps track of the next trace file number to be assigned.

  • ndb_node_id_out.log is a file containing any data output by the ndbd process. This file is created only if ndbd is started as a daemon.

  • 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.

  • ndb_node_id_signal.log 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.

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;"

See Section, “The MySQL Cluster connectstring, for additional information about this issue. Section 15.5.5, “Command Options for MySQL Cluster Processes”, describes other options for ndbd.

When ndbd starts, it actually initiates two processes. The first of these is called the “angel process”; 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 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 recommended to use several ndbd processes which belong to different node groups.

15.5.3. ndb_mgmd, the Management Server Process

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, “The MySQL Cluster connectstring, for information about using connectstrings. Section 15.5.5, “Command Options for MySQL Cluster Processes”, 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 15.4, “MySQL Cluster Configuration”, discusses how to set up this file.

  • ndb_node_id_cluster.log 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 15.6, “Management of MySQL Cluster”.

    When the size of the cluster log reaches one million bytes, the file is renamed to ndb_node_id_cluster.log.seq_id, where 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_node_id_out.log is the file used for stdout and stderr when running the management server as a daemon.

  • is the process ID file used when running the management server as a daemon.

15.5.4. ndb_mgm, the Management Client Process

The 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 1186

The default hostname and port number are localhost and 1186, respectively.

Additional information about using ndb_mgm can be found in Section, “Command Options for ndb_mgm, and Section 15.6.2, “Commands in the Management Client”.

15.5.5. Command Options for MySQL Cluster Processes

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 from those in MySQL 4.1 Cluster to make them consistent with one another as well as with mysqld. You can use the --help option to view a list of supported options.

The following sections describe options specific to individual NDB programs.

  • --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;"
  • --debug[=options]

    This option can only be used 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


    shell> ndb_mgm --execute="SHOW"

    is equivalent to

    NDB> SHOW;

    This is analogous to how the --execute or -e option works with the mysql command-line client. See Section 4.3.1, “Using Options on the Command Line”.

  • --version, -V

    Prints the version number of the ndbd process. The version number is the MySQL Cluster version number. The version number is relevant because not all versions can be used together, and at startup the MySQL Cluster processes verifies that the versions of the binaries being used can co-exist in the same cluster. This is also important when performing an online software upgrade of MySQL Cluster. MySQL Cluster-Related Command Options for mysqld

  • --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. Command Options for ndbd

For options common to NDB programs, see Section 15.5.5, “Command Options for MySQL Cluster Processes”.

  • --daemon, -d

    Instructs ndbd to execute as a daemon process. This is the default behavior. --nodaemon can be used to not start the process 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 datafiles.

    This option does not affect any backup files that have already been created by the affected node.

  • --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

    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 command to the management client. Command Options for ndb_mgmd

For options common to NDB programs, see Section 15.5.5, “Command Options for MySQL Cluster Processes”.

  • --config-file=file_name, -f file_name,

    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 file_name, but this shortcut is obsolete and should not be used in new installations.

  • --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. Command Options for ndb_mgm

For options common to NDB programs, see Section 15.5.5, “Command Options for MySQL Cluster Processes”.

  • --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.

15.6. Management of MySQL Cluster

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 15.4, “MySQL Cluster Configuration”, and Section 15.5, “Process Management in MySQL Cluster”.

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_node_id_cluster.log in the management server's DataDir directory. (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.

15.6.1. MySQL Cluster Startup Phases

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.

  • 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:

  1. Obtain a Node ID.

  2. Fetch configuration data.

  3. Allocate ports to be used for inter-node communications.

  4. Allocate memory according to settings obtained from the configuration file.

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 are established, and starts Cluster heartbeats.

  • 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 and undo logs.

    • 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:

    1. The node is included in transaction handling operations.

    2. The node schema is compared with that of the master and synchronized with it.

    3. Synchronize data received in the form of INSERT from the other data nodes in this node's node group.

    4. 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 10

    At this point in a node restart or initial node restart, APIs may connect to the node and being to receive events.

  • Stage 11

    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, transactions 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.

15.6.2. Commands in the Management Client

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 15.6.3, “Event Reports Generated in MySQL Cluster”. commands for creating backups and restoring from backup are provided in Section 15.6.5, “On-line Backup of MySQL Cluster”.

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

    Starts the data node identified by node_id (or all data nodes).

  • node_id STOP

    Stops the data node identified by node_id (or all data nodes).

  • node_id RESTART [-N] [-I]

    Restarts the data node identified by node_id (or all data nodes).

  • node_id STATUS

    Displays status information for the data node identified by node_id (or for all data nodes).


    Enters single-user mode, whereby only the MySQL server identified by the node ID node_id is allowed to access the database.


    Exits single-user mode, allowing all SQL nodes (that is, all running mysqld processes) to access the database.

  • QUIT

    Terminates the management client.


    Shuts down all cluster nodes, except for SQL nodes, and exits.

15.6.3. Event Reports Generated in MySQL 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. These are the cluster log, which includes events generated by all cluster nodes, and node logs, which are local to each data node.

Output generated by cluster event logging can have multiple destinations including a file, the management server console window, or syslog. Output generated by node event logging is written to the data node's console window.

Both types of event logs can be set to log different subsets of events.

Note: The cluster log is the log recommended for most uses because it provides logging information for an entire cluster in a single file. Node logs are intended to be used only during application development, or for debugging application code.

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 “most important” and 15 “least important.

  • 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. Logging Management Commands

The following management commands are related to the cluster log:


    Turns the cluster log on.


    Turns the cluster log off.


    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.

CategoryDefault threshold (All data nodes)

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.)

1ALERTA condition that should be corrected immediately, such as a corrupted system database
2CRITICALCritical conditions, such as device errors or insufficient resources
3ERRORConditions that should be corrected, such as configuration errors
4WARNINGConditions that are not errors, but that might require special handling
5INFOInformational messages
6DEBUGDebugging messages used for NDB Cluster development

Event severity levels can be turned on or off (using CLUSTERLOG FILTER — 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. Log Events

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 “Global Checkpoint” and “Local Checkpoint,” respectively.


These events are associated with connections between Cluster nodes.

EventPrioritySeverity LevelDescription
DB nodes connected8INFOData nodes connected
DB nodes disconnected8INFOData nodes disconnected
Communication closed8INFOSQL node or data node connection closed
Communication opened8INFOSQL node or data node connection opened


The logging messages shown here are associated with checkpoints.

EventPrioritySeverity LevelDescription
LCP stopped in calc keep GCI0ALERTLCP stopped
Local checkpoint fragment completed11INFOLCP on a fragment has been completed
Global checkpoint completed10INFOGCP finished
Global checkpoint started9INFOStart of GCP: REDO log is written to disk
Local checkpoint completed8INFOLCP completed normally
Local checkpoint started7INFOStart of LCP: data written to disk
Report undo log blocked7INFOUNDO logging blocked; buffer near overflow


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.

EventPrioritySeverity LevelDescription
Internal start signal received STTORRY15INFOBlocks received after completion of restart
Undo records executed15INFO 
New REDO log started10INFOGCI keep X, newest restorable GCI Y
New log started10INFOLog part X, start MB Y, stop MB Z
Node has been refused for inclusion in the cluster8INFONode cannot be included in cluster due to misconfiguration, inability to establish communication, or other problem
DB node neighbors8INFOShows neighboring data nodes
DB node start phase X completed4INFOA data node start phase has been completed
Node has been successfully included into the cluster3INFODisplays the node, managing node, and dynamic ID
DB node start phases initiated1INFONDB Cluster nodes starting
DB node all start phases completed1INFONDB Cluster nodes started
DB node shutdown initiated1INFOShutdown of data node has commenced
DB node shutdown aborted1INFOUnable to shut down data node normally


The following events are generated when restarting a node and relate to the success or failure of the node restart process.

EventPrioritySeverity LevelDescription
Node failure phase completed8ALERTReports completion of node failure phases
Node has failed, node state was X8ALERTReports that a node has failed
Report arbitrator results2ALERTThere are eight different possible results for arbitration attempts:
  • Arbitration check failed — less than 1/2 nodes left

  • Arbitration check succeeded — node group majority

  • Arbitration check failed — missing node group

  • Network partitioning — arbitration required

  • Arbitration succeeded — affirmative response from node X

  • Arbitration failed - negative response from node X

  • Network partitioning - no arbitrator available

  • Network partitioning - no arbitrator configured

Completed copying a fragment10INFO 
Completed copying of dictionary information8INFO 
Completed copying distribution information8INFO 
Starting to copy fragments8INFO 
Completed copying all fragments8INFO 
GCP takeover started7INFO 
GCP takeover completed7INFO 
LCP takeover started7INFO 
LCP takeover completed (state = X)7INFO 
Report whether an arbitrator is found or not6INFOThere are seven different possible outcomes when seeking an arbitrator:
  • Management server restarts arbitration thread [state=X]

  • Prepare arbitrator node X [ticket=Y]

  • Receive arbitrator node X [ticket=Y]

  • Started arbitrator node X [ticket=Y]

  • Lost arbitrator node X - process failure [state=Y]

  • Lost arbitrator node X - process exit [state=Y]

  • Lost arbitrator node X <error msg> [state=Y]


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.

EventPrioritySeverity LevelDescription
Report job scheduling statistics9INFOMean internal job scheduling statistics
Sent number of bytes9INFOMean number of bytes sent to node X
Received # of bytes9INFOMean number of bytes received from node X
Report transaction statistics8INFONumbers of: transactions, commits, reads, simple reads, writes, concurrent operations, attribute information, and aborts
Report operations8INFONumber of operations
Report table create7INFO 
Memory usage5INFOData 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.

Dead due to missed heartbeat8ALERTNode X declared “dead” due to missed heartbeat
Transporter errors2ERROR 
Transporter warnings8WARNING 
Missed heartbeats8WARNINGNode X missed heartbeat #Y
General warning events2WARNING 

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.

Sent heartbeat12INFOHeartbeat sent to node X
Create log bytes11INFOLog part, log file, MB
General information events2INFO 

15.6.4. Single-User Mode

Single-user mode allows the database administrator to restrict access to the database system to a single MySQL server (SQL node). When entering single-user mode, all connections to all other MySQL servers are closed gracefully and all running transactions are aborted. No new transactions are allowed to be started.

Once the cluster has entered single-user mode, only the designated SQL node is granted access to the database.

You can use the ALL STATUS command to see when the cluster has entered single-user mode.



After this command has executed and the cluster has entered single-user mode, the SQL node whose node ID is 5 becomes the cluster's only permitted user.

The node specified in the preceding command must be a MySQL Server node; An attempt 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. MySQL Servers waiting for a connection (that is, for the cluster to become ready and available), are again permitted to connect. The MySQL Server denoted as the single-user SQL node continues to run (if still connected) during and after the state change.



There are two recommended ways to handle a node failure when running in single-user mode:

  • Method 1:

    1. Finish all single-user mode transactions

    2. Issue the EXIT SINGLE USER MODE command

    3. Restart the cluster's data nodes

  • Method 2:

    Restart database nodes prior to entering single-user mode.

15.6.5. On-line Backup of MySQL Cluster

This section describes how to create a backup and how to restore the database from a backup at a later time. Cluster Backup Concepts

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.


    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. Using The Management Server to Create a Backup

Before starting a backup, make sure that the cluster is properly configured for performing one. (See Section, “Configuration for Cluster Backup”.)

Creating a backup using the management server involves the following steps:

  1. Start the management server (ndb_mgm).

  2. Execute the command START BACKUP.

  3. The management server will reply with the message Start of backup ordered. This means that the management server has submitted the request to the cluster, but has not yet received any response.

  4. The management server will reply Backup backup_id started, where 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.) This means that the cluster has received and processed the backup request. It does not mean that the backup has finished.

  5. The management server will signal that the backup is finished with the message Backup backup_id completed.

To abort a backup that is already in progress:

  1. Start the management server.

  2. Execute the command ABORT BACKUP backup_id. The number backup_id is the identifier of the backup that was included in the response of the management server when the backup was started (in the message Backup backup_id started).

  3. The management server will acknowledge the abort request with Abort of backup backup_id ordered; note that it has received no actual response to this request yet.

  4. After the backup has been aborted, the management server will report Backup backup_id has been aborted for reason XYZ. This means that the cluster has terminated the backup and that all files related to this backup have been removed from the cluster filesystem.

It is also possible to abort a backup in progress from the 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 it is aborted, the management server makes no explicit response. However, the fact that an invalid abort command was sent is indicated in the cluster log. How to Restore a Cluster Backup

The cluster restoration program is implemented as a separate command-line utility ndb_restore, which reads the files created by the backup and inserts the stored information into the database. The restore program must be executed once for each set of backup files. That is, as many times as there were database nodes running when the backup was created.

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. (Note that the cluster should have an empty database when starting to restore a backup.) The restore program acts as an API to the cluster and therefore requires a free connection to connect to the cluster. This can be verified with the ndb_mgm command SHOW (you can accomplish this from a system shell using ndb_mgm -e SHOW). The -c connectstring option may be used to locate the MGM node (see Section, “The MySQL Cluster connectstring, for information on connectstrings). The backup files must be present in the directory given as an argument to the restoration program.

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 — once for each database node in the cluster where the backup was taken.

Note: For rapid restoration, the data may be restored in parallel, provided that there is a sufficient number of cluster connections available. However, the data files must always be applied before the logs. Configuration for Cluster Backup

Four 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 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 Section 15.4, “MySQL Cluster Configuration”. Backup Troubleshooting

If an error code is returned when issuing a backup request, the most likely cause is insufficient memory or insufficient disk space. You should check that there is enough memory allocated for the backup. Also check that there is enough 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 “hot”, restoring a MySQL Cluster from backup is not a 100% “hot” 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.

15.7. Using High-Speed Interconnects with MySQL Cluster

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.

Currently, the MySQL Cluster codebase includes support for four different transporters. Most users today employ TCP/IP over Ethernet because it is ubiquitous. TCP/IP is also by far the best-tested transporter in MySQL Cluster.

We are working to make sure that communication with the ndbd process is made in “chunks” 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.

15.7.1. Configuring MySQL Cluster to use SCI Sockets

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.


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

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-<operating-system>-time-date. It is now time to install the package in the proper place. In this example we will place the installation in /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 – 60), the next 64 node IDs (68 – 124) are assigned to the next port, and so on, with node IDs 452 – 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 “safe” 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   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> ./

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, “MySQL Cluster SCI Transport Connections”).

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/
bash-shell> ndbd

In a tcsh environment the same thing can be accomplished with:

tcsh-shell> setenv LD_PRELOAD=/opt/DIS/lib/
tcsh-shell> ndbd

Note: MySQL Cluster can use only the kernel variant of SCI Sockets.

15.7.2. Understanding the Impact of Cluster Interconnects

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 TypeTCP/IP SocketsSCI Socket
Serial pk access400 microseconds160 microseconds
Batched pk access28 microseconds22 microseconds
Serial uk access500 microseconds250 microseconds
Batched uk access70 microseconds36 microseconds
Indexed eq-bound1250 microseconds750 microseconds
Index range24 microseconds12 microseconds

We also performed another set of tests to check the performance of SCI Sockets vis-à-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 10–70% 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 15.7.1, “Configuring MySQL Cluster to use SCI Sockets” for information on how to set up SCI sockets using ordinary TCP/IP for MySQL Cluster.

15.8. Known Limitations of MySQL Cluster

In this section, we provide a list of known limitations in MySQL Cluster releases in the 5.0.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 5.0; however, we will attempt to supply fixes for these issues in subsequent release series. If you check the “Cluster” category in the MySQL bugs database at, you can find known bugs which (if marked “5.0”) we intend to correct in upcoming releases of MySQL 5.0.

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 1.8, “How to Report Bugs or Problems”. If we do not plan to fix the problem in MySQL 5.0, we will add it to the list.

(Note: See the end of this section for a list of issues in MySQL 4.1 Cluster that have been resolved in the current version.)

  • Noncompliance in syntax (resulting in errors when running existing applications):

    • Text indexes are not supported.

    • A BIT column cannot be a primary key or part of a composite primary key.

    • Geometry datatypes (WKT and WKB) are not supported by the NDB storage engine prior to MySQL 5.0.16. (Note that spatial indexes are still not supported in MySQL 5.0.16 and newer.)

    • INSERT IGNORE is 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. (Bug #17431)

  • Non-compliance in limits or behavior (may result in errors when running existing applications):

    • Error Reporting:

      • A duplicate key error returns the error message ERROR 23000: Can't write; duplicate key in table 'tbl_name'.

    • 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, TEXT, or VARCHAR 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.

    • 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 15.4.4, “Configuration File”. Most configuration parameters can be upgraded online. These hard limits include:

      • Database memory size and index memory size (DataMemory and IndexMemory, respectively).

      • 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.)

    • All Cluster table rows are of fixed length. This means (for example) that if a table has one or more VARCHAR fields containing only relatively small values, more memory and disk space is required when using the NDB storage engine than would be the case for the same table and data using the MyISAM engine. (In other words, in the case of a VARCHAR column, the column requires the same amount of storage as a CHAR column of the same size.)

    • The maximum number of tables in a Cluster database is limited to 1792.

    • The maximum number of attributes per table is limited to 128.

    • The maximum permitted size of any one row is 8KB, not including data stored in BLOB columns.

    • 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.

  • 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 not supported, resulting in non-optimal query plans in some cases. Employ USE INDEX or FORCE INDEX as a workaround.

    • Unique hash indexes created with USING HASH cannot be used for accessing a table if NULL is given as part of the key.

    • MySQL Cluster does not support durable commits on disk. Commits are replicated, but there is no guarantee that logs are flushed to disk on commit.

  • Missing features:

    • The only supported isolation level is READ COMMITTED. (InnoDB supports READ COMMITTED, READ COMMITTED, REPEATABLE READ, and SERIALIZABLE.) See Section, “Backup Troubleshooting”, for information on how this can effect 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).

    • MySQL replication will not work correctly if updates are done on multiple MySQL servers. However, if the database partitioning scheme is done at the application level and no transactions take place across these partitions, replication can be made to work.

    • Autodiscovery of databases is not supported for multiple MySQL servers accessing the same MySQL Cluster. However, autodiscovery of tables is supported in such cases. What this means is that after a database named db_name is created or imported using one MySQL server, you should issue a CREATE DATABASE db_name statement on each additional MySQL server that accesses the same MySQL Cluster. (As of MySQL 5.0.2, you may also use CREATE SCHEMA db_name.) Once this has been done for a given MySQL server, that server should be able to detect the database tables without error.

  • 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.

    • It is not possible to make online schema changes such as those accomplished using ALTER TABLE or CREATE INDEX, as the NDB Cluster does not support autodiscovery of such changes. (However, you can import or create a table that uses a different storage engine, and then convert it to NDB using ALTER TABLE tbl_name ENGINE=NDBCLUSTER. In such a case, you must issue a FLUSH TABLES statement to force the cluster to pick up the change.)

    • Online adding or dropping of nodes is not possible (the cluster must be restarted in such cases).

    • 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.

      • In order that management nodes be able to see one another, you must restart all data nodes after bringing up the cluster. (See Bug #13070 for a detailed explanation.)

    • Multiple network interfaces for data nodes 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.

    • The maximum number of data nodes is 48.

    • The total maximum number of nodes in a MySQL Cluster is 63. This number includes all MySQL Servers (SQL nodes), data nodes, and management servers.

The following Cluster limitations in MySQL 4.1 have been resolved in MySQL 5.0 as shown below:

  • The NDB Cluster storage engine supports all character sets and collations available in MySQL 5.0.

  • Prior to MySQL 5.0.6, the maximum number of metadata objects possible was 1600. Beginning with 5.0.6, this limit is increased to 20320.

  • Cluster in MySQL 5.0 supports column indexes that make use of prefixes.

  • Unlike the case in MySQL 4.1, the Cluster storage engine in MySQL 5.0 supports MySQL' query cache. See Section 5.14, “The MySQL Query Cache”.

15.9. MySQL Cluster Development Roadmap

In this section, we discuss changes in the implementation of MySQL Cluster in MySQL 5.0 as compared to MySQL 4.1. We will also discuss our roadmap for further improvements to MySQL Cluster as currently planned for MySQL 5.1.

There are relatively few changes between the NDB Cluster storage engine implementations in MySQL 4.1 and in 5.0, so the upgrade path should be relatively quick and painless.

All significantly new features being developed for MySQL Cluster are going into the MySQL 5.1 tree. We also provide some hints about what Cluster in MySQL 5.1 is likely to include later in this section (see Section 15.9.2, “MySQL 5.1 Development Roadmap for MySQL Cluster”).

15.9.1. MySQL Cluster Changes in MySQL 5.0

MySQL Cluster in versions 5.0.3-beta and later contains a number of new features that are likely to be of interest:

  • Push-Down Conditions: Consider the following query:

    SELECT * FROM t1 WHERE non_indexed_attribute = 1;

    This query will use a full table scan and the condition will be evaluated in the cluster's data nodes. Thus, it is not necessary to send the records across the network for evaluation. (That is, function transport is used, rather than data transport.) Please note that this feature is currently disabled by default (pending more thorough testing), but it should work in most cases. This feature can be enabled through the use of the SET engine_condition_pushdown = On statement. Alternatively, you can run mysqld with the this feature enabled by starting the MySQL server with the --engine-condition-pushdown option.

    A major benefit of this change is that queries can be executed in parallel. This means that queries against non-indexed columns can run faster than previously by a factor of as much as 5 to 10 times, times the number of data nodes, because multiple CPUs can work on the query in parallel.

    You can use EXPLAIN to determine when condition pushdown is being used. See Section 7.2.1, “Optimizing Queries with EXPLAIN.

  • Decreased IndexMemory Usage: In MySQL 5.0, each record consumes approximately 25 bytes of index memory, and every unique index uses 25 bytes per record of index memory (in addition to some data memory because these are stored in a separate table). This is because the primary key is not stored in the index memory anymore.

  • Query Cache Enabled for MySQL Cluster: See Section 5.14, “The MySQL Query Cache”, for information on configuring and using the query cache.

  • New Optimizations: One optimization that merits particular attention is that a batched read interface is now used in some queries. For example, consider the following query:

    SELECT * FROM t1 WHERE primary_key IN (1,2,3,4,5,6,7,8,9,10);

    This query will be executed 2 to 3 times more quickly than in previous MySQL Cluster versions due to the fact that all 10 key lookups are sent in a single batch rather than one at a time.

  • Limit On Number of Metadata Objects: In MySQL 4.1, each Cluster database may contain a maximum of 1600 metadataobjects, including database tables, system tables, indexes and BLOBs. Beginning with MySQL 5.0.6, this number is increased to 20320.

15.9.2. MySQL 5.1 Development Roadmap for MySQL Cluster

What is said here is a status report based on recent commits to the MySQL 5.1 source tree. It should be noted all 5.1 development is subject to change.

There are currently 4 major new features being developed for MySQL 5.1:

  1. Integration of MySQL Cluster into MySQL replication: This will make 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.

  2. Support for disk-based records: Records on disk will be supported. Indexed fields including the primary key hash index must still be stored in RAM but all other fields can be on disk.

  3. 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.

  4. User-defined partitioning: Users will be able to define partitions based on columns that are part of the primary key. The MySQL Server will be able to discover whether it is possible to prune away some of the partitions from the WHERE clause. Partitioning based on KEY, HASH, RANGE, and LIST handlers will be possible, as well as subpartitioning. This feature should also be available for many other handlers, and not only NDB Cluster.

In addition, we are working to increase the 8KB size limit for rows containing columns of types other than BLOB or TEXT in Cluster tables. This is due to the fact that rows are currently fixed in size and the page size is 32,768 bytes (minus 128 bytes for the row header). Currently, this means that if we allowed more than 8KB per record, any remaining space (up to approximately 14,000 bytes) would be left empty. In MySQL 5.1, we plan to fix this limitation so that using more than 8KB in a given row does not result in the remainder of the page being wasted.

15.10. MySQL Cluster FAQ

This section answers questions that are often asked about MySQL Cluster.

  • What does “NDB” mean?

    This stands for “Network Database.

  • What's the difference in using Cluster vs. using replication?

    In a replication setup, a master MySQL server updates one or more slaves. Transactions are committed sequentially, and a slow transaction can cause the slave to lag behind the master. This means that if the master fails, it is possible that the slave might not have recorded the last few transactions. If a transaction-safe engine such as InnoDB is being used, a transaction will either be complete on the slave or not applied at all, but replication does not guarantee that all data on the master and the slave will be consistent at all times. In MySQL Cluster, all data nodes are kept in synchrony, and a transaction committed by any one data node is committed for all data nodes. In the event of a data node failure, all remaining data nodes remain in a consistent state.

    In short, whereas standard MySQL replication is asynchronous, MySQL Cluster is synchronous.

    We are planning to implement (asynchronous) replication for Cluster in MySQL 5.1. This will include the capability to replicate both between two clusters and between a MySQL cluster and a non-Cluster MySQL server.

  • Do I need to do any special networking to run Cluster? (How do computers in a cluster communicate?)

    MySQL Cluster is intended to be used in a high-bandwidth environment, with computers connecting via TCP/IP. Its performance depends directly upon the connection speed between the cluster's computers. The minimum connectivity requirements for Cluster include a typical 100-megabit Ethernet network or the equivalent. We recommend you use gigabit Ethernet whenever available.

    The faster SCI protocol is also supported, but requires special hardware. See Section 15.7, “Using High-Speed Interconnects with MySQL Cluster”, for more information about SCI.

  • How many computers do I need to run a cluster, and why?

    A minimum of three computers is required to run a viable cluster. However, the minimum recommended number of computers in a MySQL Cluster is four: one each to run the management and SQL nodes, and two computers to serve as storage nodes. The purpose of the two data nodes is to provide redundancy; the management node must run on a separate machine to guarantee continued arbitration services in the event that one of the data nodes fails.

  • What do the different computers do in a cluster?

    A MySQL Cluster has both a physical and logical organization, with computers being the physical elements. The logical or functional elements of a cluster are referred to as nodes, and a computer housing a cluster node is sometimes referred to as a cluster host. Ideally, there will be one node per cluster host, although it is possible to run multiple nodes on a single host. There are three types of nodes, each corresponding to a specific role within the cluster. These are:

    • Management node (MGM node): Provides management services for the cluster as a whole, including startup, shutdown, backups, and configuration data for the other nodes. The management node server is implemented as the application ndb_mgmd; the management client used to control MySQL Cluster via the MGM node is ndb_mgm.

    • Data node: Stores and replicates data. Data node functionality is handled by an instance of the NDB data node process ndbd.

    • SQL node: This is simply an instance of MySQL Server (mysqld) that is built with support for the NDB Cluster storage engine and started with the --ndb-cluster option to enable the engine.

  • With which operating systems can I use Cluster?

    MySQL Cluster is officially supported on Linux, Mac OS X, and Solaris. We are working to add Cluster support for other platforms, including Windows, and our goal is eventually to offer MySQL Cluster on all platforms for which MySQL itself is supported.

    It may be possible to run Cluster processes on other operating systems. We have had reports from users who say that they have run Cluster successfully on FreeBSD. However, Cluster on any but the three platforms mentioned here should be considered alpha software (at best), cannot be guaranteed reliable in a production setting, and is not supported by MySQL AB.

  • What are the hardware requirements for running MySQL Cluster?

    Cluster should run on any platform for which NDB-enabled binaries are available. Naturally, faster CPUs and more memory will improve performance, and 64-bit CPUs will likely be more effective than 32-bit processors. There must be sufficient memory on machines used for data nodes to hold each node's share of the database (see How much RAM do I Need? for more information). Nodes can communicate via a standard TCP/IP network and hardware. For SCI support, special networking hardware is required.

  • How much RAM do I need? Is it possible to use disk memory at all?

    In MySQL-5.0, Cluster is in-memory only. This means that all table data (including indexes) is stored in RAM. Therefore, if your data takes up 1GB of space and you want to replicate it once in the cluster, you need 2GB of memory to do so. This in addition to the memory required by the operating system and any applications running on the cluster computers.

    If a data node's memory usage exceeds what is available in RAM, then the system will attempt to use swap space up to the limit set for DataMemory. However, this will at best result in severely degraded performance, and may cuase the node to be dropped due to slow response time (missed hearbeats). We do not recommend on relying on disk swapping in a production environment for this reason. In any case, once the DataMemory limit is reached, any operations requiring additional memory (such as inserts) will fail.

    We are working to implement disk data storage for MySQL Cluster in a future MySQL release series, but we have no plans to add this capability in MySQL 5.0.

    You can use the following formula for obtaining a rough estimate of how much RAM is needed for each data node in the cluster:

    (SizeofDatabase × NumberOfReplicas × 1.1 ) / NumberOfDataNodes

    To calculate the memory requirements more exactly requires determining, for each table in the cluster database, the storage space required per row (see Section 11.5, “Data Type Storage Requirements”, for details), and multiplying this by the number of rows. You must also remember to account for any column indexes as follows:

    • Each primary key or hash index created for an NDBCluster table requires 21–25 bytes per record. These indexes use IndexMemory.

    • Each ordered index requires 10 bytes storage per record, using DataMemory.

    • Creating a primary key or unique index also creates an ordered index, unless this index is created with USING HASH. In other words, if created without USING HASH, a primary key or unique index on a Cluster table takes up 31–35 bytes per record in MySQL 5.0.

      Note that creating MySQL Cluster tables with USING HASH for all primary keys and unique indexes will generally cause table updates to run more quickly. This is due to the fact that less memory is required (because no ordered indexes are created), and that less CPU must be utilized (because fewer indexes must be read and possibly updated).

    It is especially important to keep in mind that every MySQL Cluster table must have a primary key. The NDB storage engine creates a primary key automatically if none is defined, and this primary key is created without USING HASH.

    There is no easy way to determine exactly how much memory is being used for storage of Cluster indexes at any given time; however, warnings are written to the Cluster log when 80% of available DataMemory or IndexMemory is in use, and again when use reaches 85%, 90%, and so on.

    We often see questions from users who report that, when they are trying to populate a Cluster database, the loading process terminates prematurely and an error message like this one is observed:

    ERROR 1114: The table 'my_cluster_table' is full

    When this occurs, the cause is very likely to be that your setup does not provide sufficient RAM for all table data and all indexes, including the primary key required by the NDB storage engine and automatically created in the event that the table definition does not include the definition of a primary key.

    It is also worth noting that all data nodes should have the same amount of RAM, as no data node in a cluster can use more memory than the least amount available to any individual data node. In other words, if there are three computers hosting Cluster data nodes, with two of these having 3GB of RAM available to store Cluster data, and one having only 1GB RAM, then each data node can devote only 1GB to clustering.

  • Because MySQL Cluster uses TCP/IP, does that mean I can run it over the Internet, with one or more nodes in a remote location?

    It is very doubtful in any case that a cluster would perform reliably under such conditions, as MySQL Cluster was designed and implemented with the assumption that it would be run under conditions guaranteeing dedicated high-speed connectivity such as that found in a LAN setting using 100 Mbps or gigabit Ethernet (preferably the latter). We neither test nor warrant its performance using anything slower than this.

    Also, it is extremely important to keep in mind that communications between the nodes in a MySQL Cluster are not secure; they are neither encrypted nor safeguarded by any other protective mechanism. The most secure configuration for a cluster is in a private network behind a firewall, with no direct access to any Cluster data or management nodes from outside. (For SQL nodes, you should take the same precautions as you would with any other instance of the MySQL server.)

  • Do I have to learn a new programming or query language to use Cluster?

    No. Although some specialized commands are used to manage and configure the cluster itself, only standard (My)SQL queries and commands are required for the following operations:

    • Creating, altering, and dropping tables

    • Inserting, updating, and deleting table data

    • Creating, changing, and dropping primary and unique indexes

    • Configuring and managing SQL nodes (MySQL servers)

  • How do I find out what an error or warning message means when using Cluster?

    There are two ways in which this can be done:

    • From within the mysql client, use SHOW ERRORS or SHOW WARNINGS immediately upon being notified of the error or warning condition. Errors and warnings also be displayed in MySQL Query Browser.

    • From a system shell prompt, use perror --ndb error_code.

  • Is MySQL Cluster transaction-safe? What isolation levels are supported?

    Yes: For tables created with the NDB storage engine, transactions are supported. In MySQL 5.0, Cluster supports only the READ COMMITTED transaction isolation level.

  • What storage engines are supported by MySQL Cluster?

    Clustering in MySQL is supported only by the NDB storage engine. That is, in order for a table to be shared between nodes in a cluster, it must be created using ENGINE=NDB (or ENGINE=NDBCLUSTER, which is equivalent).

    (It is possible to create tables using other storage engines such as MyISAM or InnoDB on a MySQL server being used for clustering, but these non-NDB tables will not participate in the cluster.)

  • Which versions of the MySQL software support Cluster? Do I have to compile from source?

    Cluster is supported in all MySQL-max binaries in the 5.0 release series, except as noted in the following paragraph. You can determine whether your server has NDB support using either the SHOW VARIABLES LIKE 'have_%' or SHOW ENGINES statement. (See Section 5.3, “The mysqld-max Extended MySQL Server”, for more information.)

    Linux users, please note that NDB is not included in the standard MySQL server RPMs. Beginning with MySQL 5.0.4, there are separate RPM packages for the NDB storage engine and accompanying management and other tools; see the NDB RPM Downloads section of the MySQL 5.0 Downloads page for these. (Prior to 5.0.4, you had to use the -max binaries supplied as .tar.gz archives. This is still possible, but is not required, so you can use your Linux distribution's RPM manager if you prefer.) You can also obtain NDB support by compiling the -max binaries from source, but it is not necessary to do so simply to use MySQL Cluster. To download the latest binary, RPM, or source distribution in the MySQL 5.0 series, visit

  • In the event of a catastrophic failure — say, for instance, the whole city loses power and my UPS fails — would I lose all my data?

    All committed transactions are logged. Therefore, although it is possible that some data could be lost in the event of a catastrophe, this should be quite limited. Data loss can be further reduced by minimizing the number of operations per transaction. (It is not a good idea to perform large numbers of operations per transaction in any case.)

  • Is it possible to use FULLTEXT indexes with Cluster?

    FULLTEXT indexing is not currently supported by the NDB storage engine, or by any storage engine other than MyISAM. We are working to add this capability in a future release.

  • Can I run multiple nodes on a single computer?

    It is possible but not advisable. One of the chief reasons to run a cluster is to provide redundancy. To enjoy the full benefits of this redundancy, each node should reside on a separate machine. If you place multiple nodes on a single machine and that machine fails, you lose all of those nodes. Given that MySQL Cluster can be run on commodity hardware loaded with a low-cost (or even no-cost) operating system, the expense of an extra machine or two is well worth it to safeguard mission-critical data. It also worth noting that the requirements for a cluster host running a management node are minimal. This task can be accomplished with a 200 MHz Pentium CPU and sufficient RAM for the operating system plus a small amount of overhead for the ndb_mgmd and ndb_mgm processes.

  • Can I add nodes to a cluster without restarting it?

    Not at present. A simple restart is all that is required for adding new MGM or SQL nodes to a Cluster. When adding data nodes the process is more complex, and requires the following steps:

    1. Make a complete backup of all Cluster data.

    2. Completely shut down the cluster and all cluster node processes.

    3. Restart the cluster, using the --initial startup option.

    4. Restore all cluster data from the backup.

    In a future MySQL Cluster release series, we hope to implement a “hot” reconfiguration capability for MySQL Cluster to minimize (if not eliminate) the requirement for restarting the cluster when adding new nodes.

  • Are there any limitations that I should be aware of when using Cluster?

    NDB tables in MySQL are subject to the following limitations:

    • Not all character sets and collations are supported.

    • FULLTEXT indexes and index prefixes are not supported. Only complete columns may be indexed.

    • Spatial data types are not supported. See Chapter 16, Spatial Extensions.

    • Only complete rollbacks for transactions are supported. Partial rollbacks and rollbacks to savepoints are not supported.

    • The maximum number of attributes allowed per table is 128, and attribute names cannot be any longer than 31 characters. For each table, the maximum combined length of the table and database names is 122 characters.

    • The maximum size for a table row is 8 kilobytes, not counting BLOBs. There is no set limit for the number of rows per table. Table size limits depend on a number of factors, in particular on the amount of RAM available to each data node.

    • The NDB engine does not support foreign key constraints. As with MyISAM tables, these are ignored.

    • Query caching is not supported.

    For additional information on Cluster limitations, see Section 15.8, “Known Limitations of MySQL Cluster”.

  • How do I import an existing MySQL database into a cluster?

    You can import databases into MySQL Cluster much as you would with any other version of MySQL. Other than the limitation mentioned in the previous question, the only other special requirement is that any tables to be included in the cluster must use the NDB storage engine. This means that the tables must be created with ENGINE=NDB or ENGINE=NDBCLUSTER. It is also possible to convert existing tables using other storage engines to NDB Cluster using ALTER TABLE, but requires an additional workaround. See Section 15.8, “Known Limitations of MySQL Cluster”, for details.

  • How do cluster nodes communicate with one another?

    Cluster nodes can communicate via any of three different protocols: TCP/IP, SHM (shared memory), and SCI (Scalable Coherent Interface). Where available, SHM is used by default between nodes residing on the same cluster host. SCI is a high-speed (1 gigabit per second and higher), high-availability protocol used in building scalable multi-processor systems; it requires special hardware and drivers. See Section 15.7, “Using High-Speed Interconnects with MySQL Cluster”, for more about using SCI as a transport mechanism in MySQL Cluster.

  • What is an “arbitrator”?

    If one or more nodes in a cluster fail, it is possible that not all cluster nodes will be able to “see” one another. In fact, it is possible that two sets of nodes might become isolated from one another in a network partitioning, also known as a “split brain” scenario. This type of situation is undesirable because each set of nodes tries to behave as though it is the entire cluster.

    When cluster nodes go down, there are two possibilities. If more than 50% of the remaining nodes can communicate with each other, we have what is sometimes called a “majority rules” situation, and this set of nodes is considered to be the cluster. The arbitrator comes into play when there is an even number of nodes: in such cases, the set of nodes to which the arbitrator belongs is considered to be the cluster, and nodes not belonging to this set are shut down.

    The preceding information is somewhat simplified. A more complete explanation taking into account node groups follows:

    When all nodes in at least one node group are alive, network partitioning is not an issue, because no one portion of the cluster can form a functional cluster. The real problem arises when no single node group has all its nodes alive, in which case network partitioning (the “split-brain” scenario) becomes possible. Then an arbitrator is required. All cluster nodes recognize the same node as the arbitrator, which is normally the management server; however, it is possible to configure any of the MySQL Servers in the cluster to act as the arbitrator instead. The arbitrator accepts the first set of cluster nodes to contact it, and tells the remaining set to shut down. Arbitrator selection is controlled by the ArbitrationRank configuration parameter for MySQL Server and management server nodes. (See Section, “Defining the MySQL Cluster Management Server”, for details.) It should also be noted that the role of arbitrator does not in and of itself impose any heavy demands upon the host so designated, and thus the arbitrator host does not need to be particularly fast or to have extra memory especially for this purpose.

  • What data types are supported by MySQL Cluster?

    MySQL Cluster supports all of the usual MySQL data types, with the exception of those associated with MySQL's spatial extensions. (See Chapter 16, Spatial Extensions.) In addition, there are some differences with regard to indexes when used with NDB tables. Note: MySQL Cluster tables (that is, tables created with ENGINE=NDBCLUSTER) have only fixed-width rows. This means that (for example) each record containing a VARCHAR(255) column will require space for 255 characters (as required for the character set and collation being used for the table), regardless of the actual number of characters stored therein. This issue is expected to be fixed in a future MySQL release series.

    See Section 15.8, “Known Limitations of MySQL Cluster”, for more information about these issues.

  • How do I start and stop MySQL Cluster?

    It is necessary to start each node in the cluster separately, in the following order:

    1. Start the management node with the ndb_mgmd command.

    2. Start each data node with the ndbd command.

    3. Start each MySQL server (SQL node) using mysqld_safe --user=mysql &.

    Each of these commands must be run from a system shell on the machine housing the affected node. You can verify the cluster is running by starting the MGM management client ndb_mgm on the machine housing the MGM node.

  • What happens to cluster data when the cluster is shut down?

    The data held in memory by the cluster's data nodes is written to disk, and is reloaded in memory the next time that the cluster is started.

    To shut down the cluster, enter the following command in a shell on the machine hosting the MGM node:

    shell> ndb_mgm -e shutdown

    This causes the ndb_mgm, ndb_mgm, and any ndbd processes to terminate gracefully. MySQL servers running as Cluster SQL nodes can be stopped using mysqladmin shutdown.

    For more information, see Section 15.6.2, “Commands in the Management Client”, and Section 15.3.6, “Safe Shutdown and Restart”.

  • Is it helpful to have more than one management node for a cluster?

    It can be helpful as a fail-safe. Only one MGM node controls the cluster at any given time, but it is possible to configure one MGM as primary, and one or more additional management nodes to take over in the event that the primary MGM node fails.

  • Can I mix different kinds of hardware and operating systems in a Cluster?

    Yes, so long as all machines and operating systems have the same endianness (all big-endian or all little-endian). It is also possible to use different MySQL Cluster releases on different nodes. However, we recommend this be done only as part of a rolling upgrade procedure.

  • Can I run two data nodes on a single host? Two SQL nodes?

    Yes, it is possible to do this. In the case of multiple data nodes, each node must use a different data directory. If you want to run multiple SQL nodes on one machine, each instance of mysqld must use a different TCP/IP port.

  • Can I use hostnames with MySQL Cluster?

    Yes, it is possible to use DNS and DHCP for cluster hosts. However, if your application requires “five nines” availability, we recommend using fixed IP addresses. Making communication between Cluster hosts dependent on services such as DNS and DHCP introduces additional points of failure, and the fewer of these, the better.

15.11. MySQL Cluster Glossary

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 15.4.4, “Configuration File”, 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 “node” 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 Obsolete Terms: In older versions of the MySQL Cluster documentation, data nodes were sometimes referred to as “database nodes,” “DB nodes,” or occasionally “storage nodes.” In addition, SQL nodes were sometimes known as “client nodes” or “API nodes.” This older terminology has been deprecated to minimize confusion, and for these reasons should be avoided.

  • 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.

  • 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 “partition”; however, “fragment” 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 15.7.1, “Configuring MySQL Cluster to use SCI Sockets”. For a basic introduction to SCI, see this essay at

    • 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).

  • Share-nothing architecture:

    The ideal architecture for a MySQL Cluster. In a true share-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.

  • Table:

    As is usual in the context of a relational database, the term “table” 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 15.5, “Process Management in MySQL Cluster”.

  • 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 15.6.3, “Event Reports Generated in MySQL Cluster”. 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.