New Disk Data table file format.  A new file format is used in NDB 7.6 for NDB Disk Data tables, which makes it possible for each Disk Data table to be uniquely identified without reusing any table IDs. This should help resolve issues with page and extent handling that were visible to the user as problems with rapid creating and dropping of Disk Data tables, and for which the old format did not provide a ready means to fix.

The new format is now used whenever new undo log file groups and tablespace data files are created. Files relating to existing Disk Data tables continue to use the old format until their tablespaces and undo log file groups are re-created.

To avoid problems relating to the changes in format, you should re-create any existing tablespaces and undo log file groups when upgrading to NDB 7.6. You can do this by performing an initial restart of each data node (that is, using the --initial option) as part of the upgrade process. You can expect this step to be made mandatory as part of upgrading from NDB 7.5 or an earlier release series to NDB 7.6 or later.

If you are using Disk Data tables, a downgrade from any NDB 7.6 release—without regard to release status—to any NDB 7.5 or earlier release requires that you restart all data nodes with --initial as part of the downgrade process. This is because NDB 7.5 and earlier release series are not able to read the new Disk Data file format.

For more information, see Section 21.3.7, “Upgrading and Downgrading NDB Cluster”.

Data memory pooling and dynamic index memory.  Memory required for indexes on NDB table columns is now allocated dynamically from that allocated for DataMemory. For this reason, the IndexMemory configuration parameter is now deprecated, and subject to removal in a future release series.

The pooling together of index memory with data memory simplifies the configuration of NDB; a further benefit of these changes is that scaling up by increasing the number of LDM threads is no longer limited by having set an insufficiently large value for IndexMemory.This is because index memory is no longer a static quantity which is allocated only once (when the cluster starts), but can now be allocated and deallocated as required. Previously, it was sometimes the case that increasing the number of LDM threads could lead to index memory exhaustion while large amounts of DataMemory remained available.

As part of this work, a number of instances of DataMemory usage not directly related to storage of table data now use transaction memory instead.

For this reason, it may be necessary on some systems to increase SharedGlobalMemory to allow transaction memory to increase when needed, such as when using NDB Cluster Replication, which requires a great deal of buffering on the data nodes. On systems performing initial bulk loads of data, it may be necessary to break up very large transactions into smaller parts.

In addition, data nodes now generate MemoryUsage events (see Section 21.6.3.2, “NDB Cluster Log Events”) and write appropriate messages in the cluster log when resource usage reaches 99%, as well as when it reaches 80%, 90%, or 100%, as before.

Other related changes are listed here:

ndbinfo processes and config_nodes tables.  NDB 7.6 adds two tables to the ndbinfo information database to provide information about cluster nodes; these tables are listed here:

System name.  The system name of an NDB cluster can be used to identify a specific cluster. In NDB 7.6, the MySQL Server shows this name as the value of the Ndb_system_name status variable; NDB API applications can use the Ndb_cluster_connection::get_system_name() method which is added in the same release.

A system name based on the time the management server was started is generated automatically>; you can override this value by adding a [system] section to the cluster's configuration file and setting the Name parameter to a value of your choice in this section, prior to starting the management server.

ndb_import CSV import tool.  ndb_import, added in NDB Cluster 7.6, loads CSV-formatted data directly into an NDB table using the NDB API (a MySQL server is needed only to create the table and database in which it is located). ndb_import can be regarded as an analog of mysqlimport or the LOAD DATA SQL statement, and supports many of the same or similar options for formatting of the data.

Assuming that the database and target NDB table exist, ndb_import needs only a connection to the cluster's management server (ndb_mgmd) to perform the importation; for this reason, there must be an [api] slot available to the tool in the cluster's config.ini file purpose.

See Section 21.5.14, “ndb_import — Import CSV Data Into NDB”, for more information.

ndb_top monitoring tool.  Added the ndb_top utility, which shows CPU load and usage information for an NDB data node in real time. This information can be displayed in text format, as an ASCII graph, or both. The graph can be shown in color, or using grayscale.

ndb_top connects to an NDB Cluster SQL node (that is, a MySQL Server). For this reason, the program must be able to connect as a MySQL user having the SELECT privilege on tables in the ndbinfo database.

ndb_top is available for Linux, Solaris, and macOS platforms, but is not currently available for Windows platforms.

For more information, see Section 21.5.29, “ndb_top — View CPU usage information for NDB threads”.

Code cleanup.  A significant number of debugging statements and printouts not necessary for normal operations have been moved into code used only when testing or debugging NDB, or dispensed with altogether. This removal of overhead should result in a noticeable improvement in the performance of LDM and TC threads on the order of 10% in many cases.

LDM thread and LCP improvements.  Previously, when a local data management thread experienced I/O lag, it wrote to local checkpoints more slowly. This could happen, for example, during a disk overload condition. Problems could occur because other LDM threads did not always observe this state, or do likewise. NDB now tracks I/O lag mode globally, so that this state is reported as soon as at least one thread is writing in I/O lag mode; it then makes sure that the reduced write speed for this LCP is enforced for all LDM threads for the duration of the slowdown condition. Because the reduction in write speed is now observed by other LDM instances, overall capacity is increased; this enables the disk overload (or other condition inducing I/O lag) to be overcome more quickly in such cases than it was previously.

NDB error identification.  Error messages and information can be obtained using the mysql client in NDB 7.6 from a new error_messages table in the ndbinfo information database. In addition, NDB 7.6 introduces a new command-line client ndb_perror for obtaining information from NDB error codes; this replaces using perror with --ndb, which is now deprecated and subject to removal in a future release.

For more information, see Section 21.6.15.21, “The ndbinfo error_messages Table”, and Section 21.5.17, “ndb_perror — Obtain NDB Error Message Information”.

SPJ improvements.  When executing a scan as a pushed join (that is, the root of the query is a scan), the DBTC block sends an SPJ request to a DBSPJ instance on the same node as the fragment to be scanned. Formerly, one such request was sent for each of the node's fragments. As the number of DBTC and DBSPJ instances is normally set less than the number of LDM instances, this means that all SPJ instances were involved in the execution of a single query, and, in fact, some SPJ instances could (and did) receive multiple requests from the same query. NDB 7.6 makes it possible for a single SPJ request to handle a set of root fragments to be scanned, so that only a single SPJ request (SCAN_FRAGREQ) needs to be sent to any given SPJ instance (DBSPJ block) on each node.

Since DBSPJ consumes a relatively small amount of the total CPU used when evaluating a pushed join, unlike the LDM block (which is repsonsible for the majority of the CPU usage), introducing multiple SPJ blocks adds some parallelism, but the additional overhead also increases. By enabling a single SPJ request to handle a set of root fragments to be scanned, such that only a single SPJ request is sent to each DBSPJ instance on each node and batch sizes are allocated per fragment, the multi-fragment scan can obtain a larger total batch size, allowing for some scheduling optimizations to be done within the SPJ block, which can scan a single fragment at a time (giving it the total batch size allocation), scan all fragments in parallel using smaller sub-batches, or some combination of the two.

This work is expected to increase performance of pushed-down joins for the following reasons:

Improved O_DIRECT handling for redo logs.  NDB 7.6 provides a new data node configuration parameter ODirectSyncFlag which causes completed redo log writes using O_DIRECT to be handled as fsync calls. ODirectSyncFlag is disabled by default; to enable it, set it to true.

You should bear in mind that the setting for this parameter is ignored when at least one of the following conditions is true:

Locking of CPUs to offline index build threads.  In NDB 7.6, offline index builds by default use all cores available to ndbmtd, instead of being limited to the single core reserved for the I/O thread. It also becomes possible to specify a desired set of cores to be used for I/O threads performing offline multithreaded builds of ordered indexes. This can improve restart and restore times and performance, as well as availability.

This improvement involves several related changes. The first of these is to change the default value for the BuildIndexThreads configuration parameter (from 0 to 128), means that offline ordered index builds are now multithreaded by default. The default value for the TwoPassInitialNodeRestartCopy is also changed (from false to true), so that an initial node restart first copies all data without any creation of indexes from a “live” node to the node which is being started, builds the ordered indexes offline after the data has been copied, then again synchronizes with the live node; this can significantly reduce the time required for building indexes. In addition, to facilitate explicit locking of offline index build threads to specific CPUs, a new thread type (idxbld) is defined for the ThreadConfig configuration parameter.

As part of this work, NDB can now distinguish between execution thread types and other types of threads, and between types of threads which are permanently assigned to specific tasks, and those whose assignments are merely temporary.

NDB 7.6 also introduces the nosend parameter for ThreadCOnfig. By setting this to 1, you can keep a main, ldm, rep, or tc thread from assisting the send threads. This parameter is 0 by default, and cannot be used with I/O threads, send threads, index build threads, or watchdog threads.

For additonal information, see the descriptions of the parameters.

Variable batch sizes for DDL bulk data operations.  As part of work ongoing to optimize bulk DDL performance by ndbmtd, it is now possible to obtain performance improvements by increasing the batch size for the bulk data parts of DDL operations processing data using scans. Batch sizes are now made configurable for unique index builds, foreign key builds, and online reorganization, by setting the respective data node configuration parameters listed here:

For each of the parameters just listed, the default value is 64, the minimum is 16, and the maximum is 512.

Increasing the appropriate batch size or sizes can help amortize inter-thread and inter-node latencies and make use of more parallel resources (local and remote) to help scale DDL performance. In each case there can be a tradeoff with ongoing traffic.

Partial LCPs.  NDB 7.6 implements partial local checkpoints. Formerly, an LCP always made a copy of the entire database. When working with terabytes of data this process could require a great deal of time, with an adverse impact on node and cluster restarts especially, as well as more space for the redo logs. It is now no longer strictly necessary for LCPs to do this—instead, an LCP now by default saves only a number of records that is based on the quantity of data changed since the previous LCP. This can vary between a full checkpoint and a checkpoint that changes nothing at all. In the event that the checkpoint reflects any changes, the minimum is to write one part of the 2048 making up a local LCP.

As part of this change, two new data node configuration parameters are inroduced in this release: EnablePartialLcp (default true, or enabled) enables partial LCPs. RecoveryWork controls the percentage of space given over to LCPs; it increases with the amount of work which must be performed on LCPs during restarts as opposed to that performed during normal operations. Raising this value causes LCPs during normal operations to require writing fewer records and so decreases the usual workload. Raising this value also means that restarts can take longer.

You must disable partial LCPs explicitly by setting EnablePartialLcp=false. This uses the least amount of disk, but also tends to maximize the write load for LCPs. To optimize for the lowest workload on LCPs during normal operation, use EnablePartialLcp=true and RecoveryWork=100. To use the least disk space for partial LCPs, but with bounded writes, use EnablePartialLcp=true and RecoveryWork=25, which is the minimum for RecoveryWork. The default is EnablePartialLcp=true with RecoveryWork=50, which means LCP files require approximately 1.5 times DataMemory; using CompressedLcp=1, this can be further reduced by half. Recovery times using the default settings should also be much faster than when EnablePartialLcp is set to false.

In addition the data node configuration parameters BackupDataBufferSize, BackupWriteSize, and BackupMaxWriteSize are all now deprecated, and subject to removal in a future release of MySQL NDB Cluster.

As part of this enhancement, work has been done to correct several issues with node restarts wherein it was possible to run out of undo log in various situations, most often when restoring a node that had been down for a long time during a period of intensive write activity.

Additional work was done to improve data node survival of long periods of synchronization without timing out, by updating the LCP watchdog during this process, and keeping better track of the progress of disk data synchronization. Previously, there was the possibility of spurious warnings or even node failures if synchronization took longer than the LCP watchdog timeout.

Parallel undo log record processing.  Formerly, the data node LGMAN kernel block processed undo log records serially; now this is done in parallel. The rep thread, which hands off undo records to LDM threads, waited for an LDM to finish applying a record before fetching the next one; now the rep thread no longer waits, but proceeds immediately to the next record and LDM.

A count of the number of outstanding log records for each LDM in LGMAN is kept, and decremented whenever an LDM has completed the execution of a record. All the records belonging to a page are sent to the same LDM thread but are not guaranteed to be processed in order, so a hash map of pages that have outstanding records maintains a queue for each of these pages. When the page is available in the page cache, all records pending in the queue are applied in order.

A few types of records continue to be processed serially: UNDO_LCP, UNDO_LCP_FIRST, UNDO_LOCAL_LCP, UNDO_LOCAL_LCP_FIRST, UNDO_DROP, and UNDO_END.

There are no user-visible changes in functionality directly associated with this performance enhancement; it is part of work done to improve undo long handling in support of partial local checkpoints in NDB Cluster 7.6.

Reading table and fragment IDs from extent for undo log applier.  When applying an undo log, it is necessary to obtain the table ID and fragment ID from the page ID. This was done previously by reading the page from the PGMAN kernel block using an extra PGMAN worker thread, but when applying the undo log it was necessary to read the page again.

when using O_DIRECT this was very inefficient since the page was not cached in the OS kernel. To correct this issue, mapping from page ID to table ID and fragment ID is now done using information from the extent header the table IDs and fragment IDs for the pages used within a given extent. The extent pages are always present in the page cache, so no extra reads from disk are required for performing the mapping. In addition, the information can already be read, using existing TSMAN kernel block data structures.

See the description of the ODirect data node configuration parameter, for more information.

Shared memory transporter.  User-defined shared memory (SHM) connections between a data node and an API node on the same host computer are fully supported in NDB 7.6, and are no longer considered experimental. You can enable an explicit shared memory connection by setting the UseShm configuration parameter to 1 for the relevant data node. When explicitly defining shared memory as the connection method, it is also necessary that both the data node and the API node are identified by HostName.

Performance of SHM connections can be enhanced through setting parameters such as ShmSize, ShmSpintime, and SendBufferMemory in an [shm] or [shm default] section of the cluster configuration file (config.ini). Configuration of SHM is otherwise similar to that of the TCP transporter.

The SigNum parameter is not used in the new SHM implementation, and any settings made for it are now ignored. Section 21.4.3.12, “NDB Cluster Shared Memory Connections”, provides more information about these parameters. In addition, as part of this work, NDB code relating to the old SCI transporter has been removed.

For more information, see Section 21.4.3.12, “NDB Cluster Shared Memory Connections”.

SPJ block inner join optimization.  In NDB 7.6, the SPJ kernel block can take into account when it is evaluating a join request in which at least some of the tables are INNER-joined. This means that it can eliminate requests for row, ranges, or both as soon as it becomes known that one or more of the preceding requests did not return any results for a parent row. This saves both the data nodes and the SPJ block from having to handle requests and result rows which never take part in an INNER-joined result row.

Consider this join query, where pk is the primary key on tables t2, t3, and t4, and columns x, y, and z are nonindexed columns:

SELECT * FROM t1
  JOIN t2 ON t2.pk = t1.x
  JOIN t3 ON t3.pk = t1.y
  JOIN t4 ON t4.pk = t1.z;

Previously, this resulted in an SPJ request including a scan on table t1, and lookups on each of the tables t2, t3, and t4; these were evaluated for every row returned from t1. For these, SPJ created LQHKEYREQ requests for tables t2, t3, and t4. Now SPJ takes into consideration the requirement that, to produce any result rows, an inner join must find a match in all tables joined; as soon as no matches are found for one of the tables, any further requests to tables having the same parent or tables are now skipped.

NDB wakeup thread.  NDB uses a poll receiver to read from sockets, to execute messages from the sockets, and to wake up other threads. When making only intermittent use of a receive thread, poll ownership is given up before starting to wake up other threads, which provides some degree of parallelism in the receive thread, but, when making constant use of the receive thread, the thread can be overburdened by tasks including wakeup of other threads.

NDB 7.6 supports offloading by the receiver thread of the task of waking up other threads to a new thread that wakes up other threads on request (and otherwise simply sleeps), making it possible to improve the capacity of a single cluster connection by roughly ten to twenty percent.

Adaptive LCP control.  NDB 7.6.7 implements an adaptive LCP control mechanism which acts in response to changes in redo log space usage. By controlling LCP disk write speed, you can help protect against a number of resource-related issues, including the following:

This work includes the following changes relating to NDB configuration parameters:

This work implements control of LCP speed chiefly to minimize the risk of running out of redo log. This is done in adapative fashion, based on the amount of redo log space used, using the alert levels, with the responses taken when these levels are attained, shown here:

Raising the level also has the effect of increasing the calculated target checkpoint speed.

LCP control has the following benefits for NDB installations:

ndb_restore options.  Beginning with NDB 7.6.9, the --nodeid and --backupid options are both required when invoking ndb_restore.

Restoring by slices.  Beginning with NDB 7.6.13, it is possible to divide a backup into roughly equal portions (slices) and to restore these slices in parallel using two new options implemented for ndb_restore:

This makes it possible to employ multiple instances of ndb_restore to restore subsets of the backup in parallel, potentially reducing the amount of time required to perform the restore operation.

For more information, see the description of the ndb_restore --num-slices option.

ndb_restore: primary key schema changes.  NDB 7.6.14 (and later) supports different primary key definitions for source and target tables when restoring an NDB native backup with ndb_restore when it is run with the --allow-pk-changes option. Both increasing and decreasing the number of columns making up the original primary key are supported.

When the primary key is extended with an additional column or columns, any columns added must be defined as NOT NULL, and no values in any such columns may be changed during the time that the backup is being taken. Because some applications set all column values in a row when updating it, whether or not all values are actually changed, this can cause a restore operation to fail even if no values in the column to be added to the primary key have changed. You can override this behavior using the --ignore-extended-pk-updates option also added in NDB 7.6.14; in this case, you must ensure that no such values are changed.

A column can be removed from the table's primary key whether or not this column remains part of the table.

For more information, see the description of the --allow-pk-changes option for ndb_restore.

ndb_blob_tool enhancements.  Beginning with NDB 7.6.14, the ndb_blob_tool utility can detect missing blob parts for which inline parts exist and replace these with placeholder blob parts (consisting of space characters) of the correct length. To check whether there are missing blob parts, use the --check-missing option with this program. To replace any missing blob parts with placeholders, use the --add-missing option.

For more information, see Section 21.5.6, “ndb_blob_tool — Check and Repair BLOB and TEXT columns of NDB Cluster Tables”.

Merging backups with ndb_restore.  In some cases, it may be desirable to consolidate data originally stored in different instances of NDB Cluster (all using the same schema) into a single target NDB Cluster. This is now supported when using backups created in the ndb_mgm client (see Section 21.6.8.2, “Using The NDB Cluster Management Client to Create a Backup”) and restoring them with ndb_restore, using the --remap-column option added in NDB 7.6.14 along with --restore-data (and possibly additional compatible options as needed or desired). --remap-column can be employed to handle cases in which primary and unique key values are overlapping between source clusters, and it is necessary that they do not overlap in the target cluster, as well as to preserve other relationships between tables such as foreign keys.

--remap-column takes as its argument a string having the format db.tbl.col:fn:args, where db, tbl, and col are, respectively, the names of the database, table, and column, fn is the name of a remapping function, and args is one or more arguments to fn. There is no default value. Only offset is supported as the function name, with args as the integer offset to be applied to the value of the column when inserting it into the target table from the backup. This column must be one of INT or BIGINT; the allowed range of the offset value is the same as the signed version of that type (this allows the offset to be negative if desired).

The new option can be used multiple times in the same invocation of ndb_restore, so that you can remap to new values multiple columns of the same table, different tables, or both. The offset value does not have to be the same for all instances of the option.

In addition, two new options are provided for ndb_desc, also beginning in NDB 7.6.14:

For more information and examples, see the description of the --remap-column option.

--ndb-log-fail-terminate option.  Beginning with NDB 7.6.14, you can cause the SQL node to terminate whenever it is unable to log all row events fully. This can be done by starting mysqld with the --ndb-log-fail-terminate option.

NDB programs—NDBT dependency removal.  The dependency of a number of NDB utility programs on the NDBT library has been removed. This library is used internally for development, and is not required for normal use; its inclusion in these programs could lead to unwanted issues when testing.

Affected programs are listed here, along with the NDB versions in which the dependency was removed:

The principal effect of this change for users is that these programs no longer print NDBT_ProgramExit - status following completion of a run. Applications that depend upon such behavior should be updated to reflect the change when upgrading to the indicated versions.

Auto-Installer deprecation and removal.  The MySQL NDB Cluster Auto-Installer web-based installation tool (ndb_setup.py) is deprecated in NDB 7.6.16, and is removed in NDB 7.6.17 and later. It is no longer supported.

ndbmemcache deprecation and removal.  ndbmemcache is no longer supported. ndbmemcache was deprecated in NDB 7.6.16, and removed in NDB 7.6.17.

Node.js support removed.  Beginning with the NDB Cluster 7.6.16 release, support for Node.js by NDB 7.6 has been removed.

Support for Node.js by NDB Cluster is maintained in NDB 8.0 only.

Conversion between NULL and NOT NULL during restore operations.  Beginning with NDB 7.6.19, ndb_restore can support restoring of NULL columns as NOT NULL and the reverse, using the options listed here:

For more information, see the descriptions of the indicated ndb_restore options.

OpenSSL 3.0 support.  Beginning with NDB 7.6.27, all MySQL server and client binaries included in the NDB distribution are compiled with support for Open SSL 3.0