21.3.3.9 MySQL Server Options and Variables for NDB Cluster

--ndbcluster

The NDBCLUSTER storage engine is necessary for using NDB Cluster. If a mysqld binary includes support for the NDBCLUSTER storage engine, the engine is disabled by default. Use the --ndbcluster option to enable it. Use --skip-ndbcluster to explicitly disable the engine.

It is not necessary or desirable to use this option together with --initialize. Beginning with NDB 7.5.4, --ndbcluster is ignored (and the NDB storage engine is not enabled) if --initialize is also used. (Bug #81689, Bug #23518923)

--ndb-allow-copying-alter-table=[ON|OFF]

Let ALTER TABLE and other DDL statements use copying operations on NDB tables. Set to OFF to keep this from happening; doing so may improve performance of critical applications.

--ndb-batch-size=#

This sets the size in bytes that is used for NDB transaction batches.

--ndb-cluster-connection-pool=#

By setting this option to a value greater than 1 (the default), a mysqld process can use multiple connections to the cluster, effectively mimicking several SQL nodes. Each connection requires its own [api] or [mysqld] section in the cluster configuration (config.ini) file, and counts against the maximum number of API connections supported by the cluster.

Suppose that you have 2 cluster host computers, each running an SQL node whose mysqld process was started with --ndb-cluster-connection-pool=4; this means that the cluster must have 8 API slots available for these connections (instead of 2). All of these connections are set up when the SQL node connects to the cluster, and are allocated to threads in a round-robin fashion.

This option is useful only when running mysqld on host machines having multiple CPUs, multiple cores, or both. For best results, the value should be smaller than the total number of cores available on the host machine. Setting it to a value greater than this is likely to degrade performance severely.

--ndb-cluster-connection-pool-nodeids=list

Specifies a comma-separated list of node IDs for connections to the cluster used by an SQL node. The number of nodes in this list must be the same as the value set for the --ndb-cluster-connection-pool option.

--ndb-cluster-connection-pool-nodeids was added in NDB 7.5.0.

--ndb-blob-read-batch-bytes=bytes

This option can be used to set the size (in bytes) for batching of BLOB data reads in NDB Cluster applications. When this batch size is exceeded by the amount of BLOB data to be read within the current transaction, any pending BLOB read operations are immediately executed.

The maximum value for this option is 4294967295; the default is 65536. Setting it to 0 has the effect of disabling BLOB read batching.

--ndb-blob-write-batch-bytes=bytes

This option can be used to set the size (in bytes) for batching of BLOB data writes in NDB Cluster applications. When this batch size is exceeded by the amount of BLOB data to be written within the current transaction, any pending BLOB write operations are immediately executed.

The maximum value for this option is 4294967295; the default is 65536. Setting it to 0 has the effect of disabling BLOB write batching.

--ndb-connectstring=connection_string

When using the NDBCLUSTER storage engine, this option specifies the management server that distributes cluster configuration data. See Section 21.3.3.3, “NDB Cluster Connection Strings”, for syntax.

--ndb-default-column-format=[FIXED|DYNAMIC]

In NDB 7.5.1 and later, sets the default COLUMN_FORMAT and ROW_FORMAT for new tables (see Section 13.1.18, “CREATE TABLE Statement”).

In NDB 7.5.1, the default for this option was DYNAMIC; in NDB 7.5.4, the default was changed to FIXED to maintain backwards compatibility with older release series (Bug #24487363).

--ndb-deferred-constraints=[0|1]

Controls whether or not constraint checks on unique indexes are deferred until commit time, where such checks are supported. 0 is the default.

This option is not normally needed for operation of NDB Cluster or NDB Cluster Replication, and is intended primarily for use in testing.

--ndb-distribution=[KEYHASH|LINHASH]

Controls the default distribution method for NDB tables. Can be set to either of KEYHASH (key hashing) or LINHASH (linear hashing). KEYHASH is the default.

--ndb-log-apply-status

Causes a replica mysqld to log any updates received from its immediate source to the mysql.ndb_apply_status table in its own binary log using its own server ID rather than the server ID of the source. In a circular or chain replication setting, this allows such updates to propagate to the mysql.ndb_apply_status tables of any MySQL servers configured as replicas of the current mysqld.

In a chain replication setup, using this option allows downstream (replica) clusters to be aware of their positions relative to all of their upstream contributors (sources).

In a circular replication setup, this option causes changes to ndb_apply_status tables to complete the entire circuit, eventually propagating back to the originating NDB Cluster. This also allows a cluster acting as a source to see when its changes (epochs) have been applied to the other clusters in the circle.

This option has no effect unless the MySQL server is started with the --ndbcluster option.

--ndb-log-empty-epochs=[ON|OFF]

Causes epochs during which there were no changes to be written to the ndb_apply_status and ndb_binlog_index tables, even when log_slave_updates is enabled.

By default this option is disabled. Disabling --ndb-log-empty-epochs causes epoch transactions with no changes not to be written to the binary log, although a row is still written even for an empty epoch in ndb_binlog_index.

Because --ndb-log-empty-epochs=1 causes the size of the ndb_binlog_index table to increase independently of the size of the binary log, users should be prepared to manage the growth of this table, even if they expect the cluster to be idle a large part of the time.

--ndb-log-empty-update=[ON|OFF]

Causes updates that produced no changes to be written to the ndb_apply_status and ndb_binlog_index tables, when when log_slave_updates is enabled.

By default this option is disabled (OFF). Disabling --ndb-log-empty-update causes updates with no changes not to be written to the binary log.

--ndb-log-exclusive-reads=[0|1]

Starting the server with this option causes primary key reads to be logged with exclusive locks, which allows for NDB Cluster Replication conflict detection and resolution based on read conflicts. You can also enable and disable these locks at runtime by setting the value of the ndb_log_exclusive_reads system variable to 1 or 0, respectively. 0 (disable locking) is the default.

For more information, see Read conflict detection and resolution.

--ndb-log-fail-terminate

When this option is specified, and complete logging of all found row events is not possible, the mysqld process is terminated.

--ndb-log-orig

Log the originating server ID and epoch in the ndb_binlog_index table.

For more information, see Section 21.6.4, “NDB Cluster Replication Schema and Tables”.

--ndb-log-transaction-id

Causes a replica mysqld to write the NDB transaction ID in each row of the binary log. Such logging requires the use of the Version 2 event format for the binary log; thus, the log_bin_use_v1_row_events system variable must be disabled to use this option.

This option is not supported in mainline MySQL Server 5.7. It is required to enable NDB Cluster Replication conflict detection and resolution using the NDB$EPOCH_TRANS() function (see NDB$EPOCH_TRANS()).

The default value is FALSE.

For more information, see Section 21.6.11, “NDB Cluster Replication Conflict Resolution”.

--ndb-log-update-minimal

Log updates in a minimal fashion, by writing only the primary key values in the before image, and only the changed columns in the after image. This may cause compatibility problems if replicating to storage engines other than NDB.

--ndb-mgmd-host=host[:port]

Can be used to set the host and port number of a single management server for the program to connect to. If the program requires node IDs or references to multiple management servers (or both) in its connection information, use the --ndb-connectstring option instead.

--ndb-nodeid=#

Set this MySQL server's node ID in an NDB Cluster.

The --ndb-nodeid option overrides any node ID set with --ndb-connectstring, regardless of the order in which the two options are used.

In addition, if --ndb-nodeid is used, then either a matching node ID must be found in a [mysqld] or [api] section of config.ini, or there must be an “open” [mysqld] or [api] section in the file (that is, a section without a NodeId or Id parameter specified). This is also true if the node ID is specified as part of the connection string.

Regardless of how the node ID is determined, its is shown as the value of the global status variable Ndb_cluster_node_id in the output of SHOW STATUS, and as cluster_node_id in the connection row of the output of SHOW ENGINE NDBCLUSTER STATUS.

For more information about node IDs for NDB Cluster SQL nodes, see Section 21.3.3.7, “Defining SQL and Other API Nodes in an NDB Cluster”.

--ndb-optimization-delay=milliseconds

Set the number of milliseconds to wait between sets of rows by OPTIMIZE TABLE statements on NDB tables. The default is 10.

--ndb-transid-mysql-connection-map=state

Enables or disables the plugin that handles the ndb_transid_mysql_connection_map table in the INFORMATION_SCHEMA database. Takes one of the values ON, OFF, or FORCE. ON (the default) enables the plugin. OFF disables the plugin, which makes ndb_transid_mysql_connection_map inaccessible. FORCE keeps the MySQL Server from starting if the plugin fails to load and start.

You can see whether the ndb_transid_mysql_connection_map table plugin is running by checking the output of SHOW PLUGINS.

--ndb-wait-connected=seconds

This option sets the period of time that the MySQL server waits for connections to NDB Cluster management and data nodes to be established before accepting MySQL client connections. The time is specified in seconds. The default value is 30.

--ndb-wait-setup=seconds

This variable shows the period of time that the MySQL server waits for the NDB storage engine to complete setup before timing out and treating NDB as unavailable. The time is specified in seconds. The default value is 30.

--skip-ndbcluster

Disable the NDBCLUSTER storage engine. This is the default for binaries that were built with NDBCLUSTER storage engine support; the server allocates memory and other resources for this storage engine only if the --ndbcluster option is given explicitly. See Section 21.3.1, “Quick Test Setup of NDB Cluster”, for an example.

ndb_autoincrement_prefetch_sz

Determines the probability of gaps in an autoincremented column. Set it to 1 to minimize this. Setting it to a high value for optimization makes inserts faster, but decreases the likelihood that consecutive autoincrement numbers will be used in a batch of inserts.

This variable affects only the number of AUTO_INCREMENT IDs that are fetched between statements; within a given statement, at least 32 IDs are obtained at a time.

ndb_cache_check_time

The number of milliseconds that elapse between checks of NDB Cluster SQL nodes by the MySQL query cache. Setting this to 0 (the default and minimum value) means that the query cache checks for validation on every query.

The recommended maximum value for this variable is 1000, which means that the check is performed once per second. A larger value means that the check is performed and possibly invalidated due to updates on different SQL nodes less often. It is generally not desirable to set this to a value greater than 2000.

ndb_clear_apply_status

By the default, executing RESET SLAVE causes an NDB Cluster replica to purge all rows from its ndb_apply_status table. You can disable t