• Log write buffering added to the and plugins (MENT-2438)

  • The buffering is controlled by the added variables:

    • server_audit_file_buffer_size - defines the size of the buffer. The default value is 0 meaning there's no buffering at all. Setting non-zero value enables the buffering with the buffer of the specified size aligned by 8192. The maximum value is 65536.

    • server_audit_sync_log_file - flushes the buffer to the log file. While the log record is in the buffer, it cannot be seen in the log file. And if there are not many events to log, the time before records can be observed can be significant. So user can do SET GLOBAL server_audit_log_file=1 to this variable to force write the buffer to the file, to make sure he doesn't miss the recent records.

• Two new fields are available via SHOW REPLICA STATUS ()

  1. Connects_Tried, which provides the number of attempts the replica has made to connect to the primary, and

  2. Master_Retry_Count, which provides the number of times the replica will attempt to connect to a primary before giving up.

• The authentication plugin caching_sha2_password has been added, not loaded by default ()

Compatibility

• As long as no new 11.4 features are used (by CREATE TABLE or DML's) and the new variable is 0 (default) one should be able to use a 10.6 slave of 11.4 master.

Developer Experience

JSON Enhancements

• New JSON function JSON_SCHEMA_VALID to validate a JSON document against a JSON schema, as documented by the JSON Schema Draft 2020. This function can also be used in a CHECK constraint to verify that only JSON documents are stored in the database which include required items and that the values are within a given range, length, etcetera.

• New JSON functions JSON_EQUALS() and JSON_NORMALIZE() for easier comparison of two JSON documents and for normalizing JSON objects to be comparable, for example when a unique key based on JSON data is needed.

• New function JSON_OVERLAPS(), which compares JSON documents to determine if they have any key-value pairs or array elements in common.\

• New function JSON_KEY_VALUE(<json_doc>,<json_path>), which extracts key/value pairs from a JSON object. The JSON path parameter is used to only return key/value pairs for matching JSON objects.

  • Example:\

• The function JSON_KEY_VALUE() can be used as an argument to JSON_TABLE(), which allows adding the key to a result set.

• Example:

• New function JSON_ARRAY_INTERSECT(<array1>, <array2>), used to find the intersection between two JSON arrays.

  • Example:

• The new JSON function JSON_OBJECT_TO_ARRAY(<json_doc>) is used to convert all JSON objects found in a JSON document to JSON arrays where each item in the outer array represents a single key-value pair from the object.

• Example:

• Resulting arrays can be compared using JSON_ARRAY_INTERSECT():

• The new JSON function JSON_OBJECT_FILTER_KEYS(<json_doc>,<array_keys>) returns key/value pairs from a JSON string for keys defined in <array_keys>.

  • Example:

• By using JSON_ARRAY_INTERSECT() and JSON_KEY() as arguments for JSON_OBJECT_FILTER_KEYS(), a comparison of two JSON strings is possible where only the same keys are compared, not the key/value pairs.\

  • Example (only show key/value pairs of json1 where the key exists in json2):

• To define the position in a JSON array from the end to the beginning, negative indexes or last can be used as the last element of an array for an JSON array of a JSON path, where the JSON path is used as a parameter in a JSON function.

• Range notation for JSON path using the keyword to define a range of elements.

SQL Functions

• New function NATURAL_SORT_KEY() which can be used to sort strings naturally.

  • Example: A string "v10" would be sorted after a string "v9"

• New function SFORMAT() for custom formats of strings. The function uses a string including formatting options and a set of given values to generate a custom formatted string.

• New function RANDOM_BYTES() which returns a binary string of a length between 1 and 1024 bytes. This nondeterministic value is generated by the random number generator of the SSL library, so it generates an arbitrary length string of cryptographic random bytes that are suitable for cryptographic use.

• The encryption functions AES_ENCRYPT() and AES_DECRYPT() now support adding the two new parameters initialization vector (iv) and block encryption mode (mode).

• Syntax for older release series:

• New syntax:

• If no mode is provided it will be used from the new system variable block_encryption_mode

  • Example (using the mode from system variable block_encryption_mode):

  • Example (mode provided as argument):

• The new options %Z and %z can be used for the format string of the DATE_FORMAT function for adding time zone information to the date string

• %Z Time zone abbreviation

• %z Numeric time zone +hhmm or -hhmm presenting the hour and minute offset from UTC

  • Example:

• The SQL function KDF() is a key derivation function, useful for generating encryption keys from a user provided password or a passphrase. It can be used to generate encryption keys for encryption functions such as AES_ENCRYPT.

• kdf_name is "hkdf" or "pbkdf2_hmac"

• width (in bits) can be any number divisible by 8

• info is a non-secret parameter of the hkdf method, it allows to generate different encryption keys for different purposes from the same secret password

• iterations is a positive numeric parameter of the pbkdf2_hmac method, larger values make the password more difficult to brute-force.

  • Example:

• The function CONV() , which converts a number between numeric base systems, now supports conversions up to base 62. This allows conversions to encodings to capital letters A-Z, lower case letters a-z, and numbers 0-9. The old limit was 36, not including lower case letters.

  • Example:

Data Types

• New data type UUID for more efficient storage of UUIDs

• New data type INET4 to store IPv4 addresses as 4-byte binary strings. Benefits of storing IPv4 addresses in the INET4 data type are:

  • Validation of incorrect values

  • Comparisons

  • Sorting

  • Functions like CAST

• Changed default behavior TIMESTAMP field properties

  • The default for explicit_defaults_for_timestamp is set to ON resulting in removing the nonstandard behavior for TIMESTAMP fields in CREATE TABLE

Stored Routines

• Stored Functions qualifiers for IN, OUT, INOUT, and IN OUT. The qualifiers are following the syntax already used for stored procedures and take the differences for Oracle into account when using the Oracle compatibility mode (sql_mode=ORACLE).

Indexes

MariaDB Enterprise Server now supports descending indexes. Composite indexes can be used with differently ordered columns to get a significant performance boost in the corresponding ORDER BY use cases.

Backported Features

• New, Detailed Replication Lag Representation

• New Information Schema Table For Password Related Data

• GTID binlog events now include the thread ID

• Automatic SST user account management for Galera

• PARSEC authentication plugin

• Extended TIMESTAMP Maximum Value

  • For 64 bit systems the maximum value for the TIMESTAMP data type has been extended from ‘2038-01-19 03:14:07 UTC’ to ‘2106-02-07 06:28:15 UTC’ without changing the storage format.

  • System versioned tables use the maximum value of TIMESTAMP for their row_end field to indicate a currently valid value. These values should be changed changed to the new maximum value:
• Option To Limit Disk Space Used For Temporary Files And Tables

  • When internal in-memory temporary tables are reaching a memory limit, they need to be stored on disk in directories defined via the tmpdir system variable.

  • Two new system variables have been added to define the maximum storage to be used for such temporary tables and other internally created temporary files:

    • max_tmp_session_space_usage

• The Software Bill of Materials (SBOM) JSON file is generated in the downloads archive

• Vector Search capability has now been added to the MariaDB Enterprise Server 11.4 release. This explains the idea of the new vector search feature in detail ()

• Vector Embeddings

  • Vector embeddings are numerical representations [0.2, -0.5, 0.8, 0.1, ...] that capture semantic meaning or features of data in a multi-dimensional space. The embedding transforms simple to complex data such as text, images, audio, or video into a series of numbers (a vector) where similar items are positioned together in the multi-dimensional vector space.

  • For example, a word embedding of the word "dog" would be close in a vector embedding space to the word "puppy", whereas "dog" would not be close to the word "airplane". The embedding representation can be used in similarity search, recommendation systems, or more generally in traditional AI/ML systems and GenAI systems.

• New Data Type VECTOR(N)

  • With the RC release a new data type VECTOR(N) has been added to represent the vector column, where N is the number of dimensions.

    • Example:

• Conversion Functions

  • VEC_FromText() converts a text representation of the vector (a json array of numbers) to a vector (little-endian IEEE float sequence of bytes, 4 bytes per float).

    • Example:

  • VEC_ToText() converts a binary vector into a json array of numbers (floats).

• Comparison Functions

  • Comparing vectors, calculating how close they are, is the key functionality needed by an application working with vector search. Two functions exist for calculating the distance between vectors. Which one to use depends on the application and on how the vectors were generated.

  • VEC_DISTANCE_EUCLIDEAN() takes two vectors and computes the straight line (L2) Euclidean distance between two multi-dimensional points in the vector space

MariaDB Enterprise Server 11.4.8-5

MariaDB Enterprise Server 11.4.9-6

Operational Enhancements

Schema and Partitioning Management / DDL

• CONVERT PARTITION and CONVERT TABLE used with ALTER TABLE can be used to convert a partition into a table or vise versa

• Exchange a Partition or Convert a Table Without Validation

  • The process of exchanging a partition with a table or converting a table to a partition can be a very slow operation, especially for larger tables because for each new data row, the partitioning definitions need to be verified to validate that the new row should indeed be in this partition.

  • This process can now be sped up by disabling this validation. This new feature should be used with care, as it can lead to inconsistencies if the partitioning rules are not met.

  • The new addition to ALTER TABLE is:

• Syntax extension to not require PARTITION keyword in each partition definition

• Stored routines are now taking changes to metadata into account whenever the stored routine is called, done for any object a stored routine depends on.

  • In older release series a reconnect was needed before a stored routine was able to update its metadata from altered objects

  • Example:

• In the above example both calls of the stored procedure p1 would have returned the same result set with the older release series.

• Temporary tables are now included in information_schema.tables and included in SHOW TABLES and SHOW TABLE STATUS

  • Example:

• General Support of Packages for Stored Routines

  • Before MariaDB Enterprise Server 11.4, the CREATE PACKAGE feature, as well as CREATE PACKAGE BODY, were only supported with sql_mode = ORACLE. They can now be used with any SQL mode.

  • Example:

System-Versioned and Application-time Period Tables

• System-Versioned Tables can automate the creation of new HISTORY partitions partitioned by INTERVAL/LIMIT using the keyword AUTO when creating a table.

• In this case a new HISTORY partition will be created on a monthly basis, storing old versions of the table data.

• Information About Application-time Period Tables

  • New views PERIOD and KEY_PERIOD_USAGE are added to information_schema.

  • View PERIODS includes the columns

    • TABLE_CATALOG

    • TABLE_SCHEMA

    • TABLE_NAME

  • View KEY_PERIOD_USAGE includes the columns

    • CONSTRAINT_CATALOG

    • CONSTRAINT_SCHEMA

• Two new columns are added to the COLUMNS view of information_schema

  • IS_SYSTEM_TIME_PERIOD_START

  • IS_SYSTEM_TIME_PERIOD_END

Backup / Restore

• A dump of historical data for system versioned tables is now possible via the new option --as-of for mariadb-dump

• System versioned tables can now be dumped and restored by mariadb-dump

  • The new parameter mariadb-dump parameter --dump-history dumps all historical data

  • To restore from a dump file the new parameter system_versioning_insert_history needs to be enabled to allow direct inserts into ROW_START and ROW_END columns

  • The existing parameter secure_timestamp needs to be set to a value which allows changing session timestamps

• The command-line tool mariadb-dump now supports the new option --order-by-size. The new option can be used to create a dump of the tables of a database according to their size, smaller tables first.

• Parallelism for mariadb-dump

  • When mariadb-dump is used with the option -T / --tab= to produce tab-separated text-format data files per table, the new option --parallel (synonym --use-threads) can be used to use several threads in parallel to dump the table data to their .txt files. Parallelism also works if the option --single-tansaction is used.

  • The option

Character Sets / Collations

• New collations based on the Unicode Collation Algorithm (UCA) 14.0.0 have been added for the character sets utf8mb3, utf8mb4, ucs2, utf16, utf32

  • One neutral and 22 language specific collations have been added

  • Case sensitive, case insensitive, and nopad variants have been added

  • Collations (how to compare characters) are now separated from character sets (how to store characters). Collation names no longer have to include character set names, and the same collation can apply to many character sets

  • Improved contraction performance in UCA collations

  • Improved UCA collation performance for utf8mb3 and utf8mb4

• Full UNICODE support for MariaDB command-line tools on recent versions of Windows (Windows 10 1909 or later, Windows 11, Windows Server 2020 supported). The server my.ini config file is now also encoded in UTF8. Command-line client mariadb.exe uses utf8mb4 by default.

• The default collation used for a character set can now be changed globally or for a session via the new system variable character_set_collations. The default collation will be used whenever a character set is defined for a database object without defining the collation.

  • When not defining a character set the default collation is still the one defined with the system variable collation_server.

  • This is also preparatory work for changing default collations to use the UCA 14.0.0 standard. In particular, this variable will allow replication between servers with different default collations.

Performance

• Optimizations to information schema system tables

  • The information schema provides tables with metadata about stored procedures and stored routines, which are often used by third party tools, and MariaDB Connectors to retrieve details about existing routines. Previously, a high number of rows in these tables would have resulted in a performance impact. We have made a number of internal improvements to eliminate the performance impact completely for use cases, where metadata had to be queried regularly.

Optimizer

• Changes to the optimizer now allow the use of an index for a comparison of a DATE function to a constant value.

• The optimizer also has been enhanced to allow single-table UPDATE and DELETE to take advantage of semi-join optimization.

• Improved optimizer performance in a case of join with many eq_ref tables

MariaDB Enterprise Cluster (powered by Galera)

• Automatic SST User Account Management for MariaDB Enterprise Cluster

  • The State Snapshot Transfer (SST) method, needed to provide a full data copy to a new node, requires a dedicated account to access the remote server (donor) during the SST process.

  • The new MariaDB Enterprise Cluster (Galera) creates the user internally for the time of an SST, which makes the need to have an account created manually obsolete. This also removes the requirement to have a user and password provided via a configuration file. Having the user created by Galera also ensures that the needed privileges are set.

• For MariaDB Galera Cluster, configurations are set using one system variable as a semicolon separated list of options, the system variable wsrep_provider_options. MariaDB Community Server system variables are limited to a length of 2048 characters, which is not sufficient for the Galera options in some use cases, and also hard to maintain as a DBA.

  • A new plugin is available, enabled via the plugin-wsrep-provider option. The options are split into separate options, if the plugin is used. The use of the plugin is optional.

MariaDB Replication

• Incompatibility change: Replication is now using Global Transaction IDs (GTID) by default to make replicas crash safe

  • The default of CHANGE MASTER TO for master_use_gtid changes from no to slave_pos

  • A fresh slave start, a RESET SLAVE, or a CHANGE MASTER TO without the defining master_use_gtid is replicating in the GTID based mode using gtid_slave_pos as the position to start downloading transactions from the primary

• Global Limitation of Space Used by Binary Logs

  • The new system variable max_binlog_total_size (alias binlog_space_limit) enables binary log purging when the total size of all binary logs exceeds the specified threshold. The default for max_binlog_total_size is 0, meaning that there is no limit. The system variable can be changed without restarting the server.

  • The new system variable

• Index for Binary Log on GTIDs

  • An index is now created on the GTIDs of the binary log, which allows a connecting replica to find the position it should start from without the need to scan the whole binary log.

    • The new system variable binlog_gtid_index (default ON) can be used to disable the creation of indexes.

• GTID Binlog Events Now Include Thread ID

  • The thread ID and the corresponding statement can now be retrieved from binary logs.

  • The output of mariadb-dump also includes the thread ID.

• A new option 'slave_max_statement_time' for the SQL thread has been added to allow to set a maximum allowed execution time for a replicated query

  • Allows to control the maximum time a replica can be behind a primary if the primary goal is to have a small replication lack, with the cost to be out of sync. It is therefore disabled by default.

• The binary log filter options binlog-do-db, binlog-ignore-db, and binlog-row-event-max-size are now visible as system variables.

  • Example:

• New "SQL_BEFORE_GTIDS" and "SQL_AFTER_GTIDS" for START REPLICA UNTIL

  • The new options SQL_BEFORE_GTIDS and SQL_AFTER_GTIDS for START REPLICA UNTIL allow user control of whether the replica stops before or after a provided GTID state. Its syntax is:

• When providing SQL_BEFORE_GTIDS="<gtid_list>", for each domain specified in the gtid_list, the replica will execute transactions up to the GTID found, and immediately stop processing events in that domain without executing the transaction of the specified GTID.

• Once all domains have stopped, the replica will stop. Events originating from domains that are not specified in the list are not replicated.

• START SLAVE UNTIL SQL_AFTER_GTIDS="<gtid_list>" is an alias to the default behavior of START SLAVE UNTIL master_gtid_pos="<gtid_list>", the known behavior before MariaDB Enterprise Server 11.4.

• The replica will execute transactions originating from domain ids provided in the list, and will stop once all transactions provided in the UNTIL list have all been executed.

• Example:

If a primary server has a binary log consisting of the following GTIDs:

Given a fresh replica (i.e., one with an empty GTID position, @@gtid_slave_pos=") is started with SQL_BEFORE_GTIDS, i.e.,

the resulting gtid_slave_pos of the replica will be "1-1-1" because the replica will execute only events from domain 1. When it sees the transaction with sequence number 2 it immediately stops without executing it.

If the replica is started with SQL_AFTER_GTIDS, i.e.,

the resulting gtid_slave_pos of the replica will be "1-1-2" because the replica will execute only events from domain 1. But in this case it stops after executing the provided GTID.

• New, detailed replication lag representation

  • The Seconds_Behind_Master field of SHOW REPLICA STATUS can be complex and confusing, especially when parallel replication, delayed replication, or the option sql_slave_skip_counter is used. To help provide a consistent view of replication lag, three new fields have been added to the statement's output to provide specific timing information about the state of the IO and SQL threads.

    • Master_last_event_time

      • Timestamp of the last event read from the primary by the IO thread

    • Slave_last_event_time

      • Timestamp from the primary of the last event committed on the replica

    • Master_Slave_time_diff

      • The difference of the above two timestamps

Optimistic ALTER TABLE for Replicas

MariaDB-binlog

• The command line tool mariadb-binlog now supports the use of global transaction IDs (GTID) for the options start-position and stop-position. mariadb-binlog can now be used to produce results filtered by the defined GTIDs.

• The command-line tool mariadb-binlog now supports the new options --do-domain-ids, --ignore-domain-ids, and --ignore-server-ids. mariadb-binlog can now be used to produce results filtered by domain ids server ids.

Security and Access Control

• Client to Server connection now SSL Encrypted by Default

  • Using SSL/TLS has been simplified with MariaDB Enterprise Server 11.4. Before version 11.4, proper SSL configuration required multiple manual steps for the server, and all the clients connecting to it.

  • Now the client can verify the server self-signed certificate without any configuration whatsoever. The server completely automatically generates the SSL certificate and the client automatically verifies it as needed.

  • This simplification allows the server to now require SSL encrypted connections by default and to refuse unencrypted connections. Additionally, MariaDB Enterprise Server 11.4 allows users to verify SSL certificates using their fingerprints.

• New (Password Authentication using Response Signed with Elliptic Curves).

  • PARSEC improves security over old authentication plugins by introducing salted passwords, time consuming key derivation function, and a client-side scramble to ensure that man-in-the-middle attackers cannot control the client response.

  • Example on how to create a user using the new authentication plugin:

• This will result in:

• The new plugin password_reuse_check can be used to validate that a password cannot be reused. The number of days until a password can be reused can be configured via a new parameter password_reuse_check_interval

• New allowlist for MariaDB Galera Cluster node to restrict the nodes which can join a cluster to increase security .

  • A new system variable wsrep_allowlist can be used to define a list of IP addresses. Only nodes from these IP addresses can join a running

• The new SQL syntax GRANT .. TO PUBLIC can now be used to easily grant privileges to databases or tables for any user, who has access to the server.

  • SHOW GRANTS FOR PUBLIC is an enhancement to the existing SHOW GRANTS syntax to retrieve all privileges granted to public

• The fine grained privileges have been removed from the SUPER privilege.

  • The SUPER privilege is still used for some special cases, like using DES_ENCRYPT and DES_DECRYPT without an explicit key, for debug settings, and some system variables for changing them with SET GLOBAL.

• New privilege SHOW CREATE ROUTINE

  • Before MariaDB Enterprise Server 11.4 a user only could see the definition of a routine, a stored feature, or function, when either of the following was met:

    • They had been granted the SELECT privilege for the mysql.procs table

• Example with the new privilege SHOW CREATE ROUTINE:

• SHOW CREATE ROUTINE privilege can be granted globally, per schema, or on individual routines.

• Retrieve Users Privileges on a Specific Table

  • MariaDB Server provides information about privileges a user has to a table in different views in the INFORMATION_SCHEMA, split between global, schema, and table privileges. Before MariaDB Enterprise Server 11.4, there was no easy way to list all the tables a user has access to, this information has to be queried from several tables.

  • MariaDB Enterprise Server 11.4 now provides a new view privileges_by_table_by_level in the SYS schema that lists the privilege and privilege level per user, schema, and table.

  • Example:

• A new information Schema view, USERS, has been added, which DBAs can use to get insights about password related information for a user.

  • This information can be used:

    • by an application to inform a user about a password about to expire or an account which is at risk of being blocked due to the number of wrong passwords entered

    • by DBAs to query users which have been blocked because of too many invalid passwords entered

  • The new view includes the fields:

    • USER - A string including user name and host

    • PASSWORD_ERRORS - A counter with the current number of wrong passwords entered

Storage Engines

• Engine-defined attributes can now also be defined per-partition for more flexible configurations.

InnoDB

• The space occupied by freed pages within the InnoDB system tablespace can be reclaimed by adding an :#autoshrink attribute to #innodb_data_file_path#, like:

• This allows the system tablespace to be truncated after the last allocated page within it, all the way to the specified minimum size (here: 12MiB). In older release series InnoDB data files never shrink in size during normal operation. One could shrink .ibd files by rebuilding tables with OPTIMIZE TABLE, or the undo tablespace files by SET GLOBAL innodb_undo_log_truncate=ON.

• Shrink temporary InnoDB tablespaces without restart

  • Before MariaDB Enterprise Server 11.4 the only way to reclaim disk space used by temporary InnoDB tablespaces was to restart the server, as temporary tablespaces are deleted when you stop the server, and are recreated with their configured size.

  • Restarting the server is not always possible, while you still need to reclaim disk space. MariaDB Enterprise Server 11.4 can solve this problem by executing

This triggers to reclaim the disk space, but existing tables will not be removed.

• Example:

• The process to import a InnoDB tablespace is now simplified. While the old process required to first create a table and to discard the tablespace before executing ALTER TABLE IMPORT TABLESPACE, now ALTER TABLE IMPORT TABLESPACE is the only command needed.

  • Example:

• The InnoDB storage engine has gone through additional code cleanup and we have removed some configuration parameters used by the InnoDB Change Buffering feature, a feature which was disabled for earlier release series.

  • New system variables like innodb_log_file_buffering, innodb_data_file_buffering, innodb_log_file_write_through, and innodb_data_file_write_through have been added to allow better control for log files and data files. They can be set dynamically while the Server is running.

• Changes to the InnoDB redo log format to reduce write amplification, which can result in better performance.

• The system variables innodb_write_io_threads and innodb_read_io_threads are now dynamic, and their values can be changed without restarting the server

Spider

• The SPIDER storage engine now allows the use of engine-defined attributes (table options), similar to other storage engines, and more convenient than the current method of providing parameters via COMMENT for a table.

Analyzing, Tracing, and Monitoring

• Improved error reporting for INSERT that inserts multiple rows. The property ROW_NUMBER in GET DIAGNOSTICS allows retrieval of the row number that caused the error or warning.

• Implementation of JSON histograms with detailed histogram collection, used when histogram_type=JSON_HB (not the default) is set. Using JSON histograms results in more precise data statistics over string data types or when columns have highly-uneven data distribution. More precise statistics allow the optimizer to create better query plans resulting in faster queries.

• ANALYZE [FORMAT=JSON] <select> has been extended to allow analyzing a query which is currently running in another connection by running SHOW ANALYZE FORMAT=JSON for <conn_id>.

• ANALYZE FORMAT=JSON now shows the time spent in the query optimizer

  • In some cases optimizing the query can take a while. ANALYZE FORMAT=JSON now reports time as "query_optimization": { "r_total_time_ms": NNNN.NNN } in the JSON string

• SHOW EXPLAIN for <conn_id>, which returns an EXPLAIN for a query running in another connection, has been extended to return the more detailed JSON output by using the syntax SHOW EXPLAIN [FORMAT=JSON] FOR <conn_id>.

  • Syntax EXPLAIN [ FORMAT=JSON] FOR CONNECTION <conn_id> is also supported.

• New status monitoring features for MariaDB Enterprise Cluster:

  • New thread states in PROCESSLIST for MariaDB Galera Cluster allow better tracking of a session status

    • "waiting to execute in isolation"

    • "waiting for TOI DDL"

• The new value SENT_ROWS in the information schema table PROCESSLIST includes the number of rows sent by the current statement, shown in the processlist.

  • Selects with functions show the total number of rows sent by the main statement and all functions

  • Stored procedures show the total number of rows sent per stored procedure statement

  • INSERT RETURNING and DELETE RETURNING show the total number of rows sent for the returning data set

  • Example:

• The SQL Error Log Plugin can be used to log errors sent to clients for later analysis. When option sql_error_log_with_db_and_thread_info=ON is set, the log file is now also showing thread id, and the current default schema for the error.

Available Versions

• MariaDB Enterprise Server 11.4.8-5

• MariaDB Enterprise Server 11.4.7-4

• MariaDB Enterprise Server 11.4.5-3

• MariaDB Enterprise Server 11.4.4-2

See also: All MariaDB Enterprise Releases

Installation Instructions

Upgrade Instructions

• Log write buffering added to the and plugins ()

  • The buffering is controlled by the added variables:

    • server_audit_file_buffer_size - defines the size of the buffer. The default value is 0 meaning there's no buffering at all. Setting non-zero value enables the buffering with the buffer of the specified size aligned by 8192. The maximum value is 65536.

• Example:

• Segmented key cache for Aria (MENT-2361)

  • A new variable (default 1) to define the number of segments has been added. The default disables the new feature.

• Backport fixes for Vectors to MariaDB Enterprise 11.4 (MENT-2365)

  • Overflow/inf in vec_distance_euclidean

  • Sporadic segmentation faults possibly related to vector search

  • Incorrect error/docs for Vector column lengths (max = 65532

  • Unexpected ER_TABLE_EXISTS_ERROR on primary or replica upon CREATE OR REPLACE for partitioned table

  • Cannot access shared MEM_ROOT without a lock

  • Can't find record in 't1' on INSERT to Vector table

  • Create vector table failed with VECTOR INDEX when innodb_force_primary_key=on

  • Assertion when adding FK to MyISAM/Aria table with a vector index

  • UBSAN: 32801 is outside the range of representable values of type 'short'

  • Always release ctx in mhnsw_delete_all

  • Optimise dot_product by loop-unrolling by a factor of 4

  • Vector index search allocates too much memory for large ef_search

  • ALTER performs vector truncation without WARN_DATA_TRUNCATED or similar warnings/errors

  • DATA/INDEX DIRECTORY options are ignored for vector index

  • DATA/INDEX DIRECTORY handling is inconsistent

  • Assert with vector index and very long PK

  • InnoDB: Failing assertion: prebuilt->select_lock_type != LOCK_NONE || srv_read_only_mode || trx->read_view.is_open()

  • Crash on disconnect when dropped Aria table with vector key under lock

  • InnoDB assert with vector index under LOCK TABLES

  • mhnsw: support powerpc64 SIMD instructions

  • mhnsw: support aarch64 SIMD instructions

  • IGNORED attribute has no effect on vector keys

  • IMPORT TABLESPACE does not work for tables with vector, although allowed

  • Vector-related error messages worth improving when possible

  • Adding a regular index on a vector column leads to invalid table structure

  • Vector values do not survive mariadb-dump / restore

  • Server crashes in Charset::mbminlen / Item_func_vec_fromtext::val_str upon mixing vector type with string

  • Server crashes in mhnsw_read_first upon using vector key with views

  • Server crashes when checking/updating a table having vector key after enabling innodb_force_primary_key

  • ALTER TABLE re-creating vector key is no-op with non-copying alter algorithms (default)

  • Server crash in FVector::distance_to upon concurrent SELECT

  • Server crashes in Item_func_vec_distance_common::get_const_arg

  • Non-copying ALTER does not pad VECTOR column, vector search further does not work

  • Assertion bitmap_is_set(&read_partitions, next->id) failed in int partition_info::vers_set_hist_part(THD *)

• Segmented key cache for Aria (MENT-2361)

  • A new variable (default 1) to define the number of segments has been added. The default disables the new feature.

• Backport fixes for Vectors to MariaDB Enterprise 11.4 (MENT-2365)

  • Overflow/inf in vec_distance_euclidean

  • Sporadic segmentation faults possibly related to vector search

  • Incorrect error/docs for Vector column lengths (max = 65532

  • Unexpected ER_TABLE_EXISTS_ERROR on primary or replica upon CREATE OR REPLACE for partitioned table

  • Cannot access shared MEM_ROOT without a lock

  • Can't find record in 't1' on INSERT to Vector table

  • Create vector table failed with VECTOR INDEX when innodb_force_primary_key=on

  • Assertion when adding FK to MyISAM/Aria table with a vector index

  • UBSAN: 32801 is outside the range of representable values of type 'short'

  • Always release ctx in mhnsw_delete_all

  • Optimise dot_product by loop-unrolling by a factor of 4

  • Vector index search allocates too much memory for large ef_search

  • ALTER performs vector truncation without WARN_DATA_TRUNCATED or similar warnings/errors

  • DATA/INDEX DIRECTORY options are ignored for vector index

  • DATA/INDEX DIRECTORY handling is inconsistent

  • Assert with vector index and very long PK

  • InnoDB: Failing assertion: prebuilt->select_lock_type != LOCK_NONE || srv_read_only_mode || trx->read_view.is_open()

  • Crash on disconnect when dropped Aria table with vector key under lock

  • InnoDB assert with vector index under LOCK TABLES

  • mhnsw: support powerpc64 SIMD instructions

  • mhnsw: support aarch64 SIMD instructions

  • IGNORED attribute has no effect on vector keys

  • IMPORT TABLESPACE does not work for tables with vector, although allowed

  • Vector-related error messages worth improving when possible

  • Adding a regular index on a vector column leads to invalid table structure

  • Vector values do not survive mariadb-dump / restore

  • Server crashes in Charset::mbminlen / Item_func_vec_fromtext::val_str upon mixing vector type with string

  • Server crashes in mhnsw_read_first upon using vector key with views

  • Server crashes when checking/updating a table having vector key after enabling innodb_force_primary_key

  • ALTER TABLE re-creating vector key is no-op with non-copying alter algorithms (default)

  • Server crash in FVector::distance_to upon concurrent SELECT

  • Server crashes in Item_func_vec_distance_common::get_const_arg

  • Non-copying ALTER does not pad VECTOR column, vector search further does not work

  • Assertion bitmap_is_set(&read_partitions, next->id) failed in int partition_info::vers_set_hist_part(THD *)