Release Notes for MariaDB Enterprise Server 10.5.22-16

Overview

MariaDB Enterprise Server 10.5.22-16 is a maintenance release of MariaDB Enterprise Server 10.5. This release includes a variety of fixes.

MariaDB Enterprise Server 10.5.22-16 was released on 2023-09-11.

Backports

  • JSON_OVERLAPS() has been backported. (MENT-1853)

    • The JSON_OVERLAPS() function can be used to compare two JSON documents to determine if they have any key-value pairs or array elements in common.

    SELECT JSON_OVERLAPS('{"A": 1, "B": {"C":2}}', '{"A": 2, "B": {"C":2}}') AS is_overlap;
    
    +---------------------+
    | is_overlap          |
    +---------------------+
    | 1                   |
    +---------------------+
    
  • JSON_SCHEMA_VALID() has been backported. (MENT-1796)

    • The JSON_SCHEMA_VALID() function can be used 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 JSON documents are only stored in the database if they include required items and that the values are within a given range and length.

Notable Changes

  • The package for the cracklib_password_check plugin includes an SELinux policy, allowing the plugin to work with SELinux. (MDEV-18374)

  • With Optimizer Trace, a sel_arg_alloc_limit_hit record is written when MAX_SEL_ARGS is reached. (MDEV-30964)

    • MAX_SEL_ARGS is one of the limits in the optimizer that triggers short-cut of time-intensive or memory-intensive analysis for complex WHERE clauses.

    • For example:

      SELECT
        JSON_DETAILED(JSON_EXTRACT(trace, '$**.setup_range_conditions'))
        FROM information_schema.OPTIMIZER_TRACE;
      
      [
         [
             {
                 "sel_arg_alloc_limit_hit":
                 {
                    "alloced_sel_args": 16001
                 }
             }
         ]
      ]
      
  • ANALYZE FORMAT=JSON output includes InnoDB statistics. (MDEV-31577)

    • For example:

      "table": {
         "table_name": "t1",
         ...
         "r_engine_stats": {
           "pages_accessed": integer,
           "pages_updated" : integer,
           "pages_read_count" : integer,
           "pages_read_time_ms" : double_val,
           "old_rows_read" : integer,
         },
      
  • Slow query log output includes InnoDB engine information. (MDEV-31558)

    • InnoDB engine information output is enabled with --log-slow-verbosity=innodb

    • Sample output:

      # Pages_accessed: 184  Pages_read: 95  Pages_updated: 0  Undo_rows_read: 1
      # Pages_read_tim e: 17.0204  Engine_time: 248.1297
      
      • Engine_time is the time in milliseconds spent inside engine calls.

      • Page_* variables are supported for the InnoDB storage engine.

  • For the Spider storage engine, the default values and behavior of some system variables has changed: (MDEV-31524)

    • Prior to this release:

      • These variables used a value of -1 (the default) to indicate that Spider would use the default table value, and this value was not user visible.

      • Spider table parameters could set values, but these values would be overridden with the system variable value if the system variable was set to a value other than -1.

    • Starting with this release:

      • The default value for these system variables has been updated to reflect the actual default table value, rather than -1.

      • Where a value is set by a table parameter, this value overrides the system default and the value set by Spider system variable.

      • If a table parameter is not set, the Spider system variable's value is used. This behavior is unchanged.

    • See "Interface Changes" for a full list of updated default values.

Issues Fixed

Can result in data loss

  • With InnoDB storage engine, when innodb_undo_log_truncate=ON, performing a recovery can result in a corrupted undo log. (MDEV-31373)

  • With InnoDB storage engine, when database encryption is enabled only on a replica, InnoDB pages can be corrupted. (MDEV-31347, MDEV-31156)

  • With InnoDB storage engine, when innodb_undo_log_truncate=ON, backup and recovery can fail. (MDEV-31487)

  • With InnoDB storage engine, if two MariaDB Server instances using InnoDB are started with the same data directory there can be data loss. (MDEV-31568)

Can result in a hang or crash

  • With Aria storage engine, when Aria encryption is enabled and used but an encryption plugin is not loaded, the server can crash. (MDEV-26258)

  • With Galera Cluster, when wsrep_sst_donor and wsrep_cluster_address are set to NULL rather than an empty string, the server can crash. (MDEV-28433)

  • When optimizer_switch='optimize_join_buffer_size=off' is set, the server can crash. (MDEV-31348)

  • When executing a SELECT query using an index for GROUP BY and filesort, the server can crash. (MDEV-30143)

  • With MariaDB Connector/C, when the mysql_list_fields() function is called against a view, the server can crash. (MDEV-30159)

  • With Aria storage engine, changing aria_sort_buffer_size settings to huge numbers and executing INSERT/UPDATE can result in a crash. (MDEV-28054)

  • With InnoDB storage engine, when slave_parallel_mode is optimistic and slave_parallel_threads is greater than 0, an ALTER SEQUENCE can fail with an out-of-order binlog error if the SEQUENCE uses InnoDB. (MDEV-31503)

    • Prior to this release, the following error can be raised:

      Last_Error: Error 'An attempt was made to binlog GTID 0-1-100 which would create an out-of-order sequence number with existing GTID 0-1-100 and gtid stric mode is enabled' on query. Default database: 'test'. Query: 'alter sequence s1 restart with 1' will be shown.

  • With SQL_MODE=ORACLE, a memory leak can occur on shutdown. (MDEV-26186)

  • Queries in the form SELECT DISTINCT function_returning_string(aggregate_func(...)) ... GROUP BY ... could produce an assertion failure or wrong result. (MDEV-31743)

  • With InnoDB storage engine, InnoDB can purge history too early, causing crash on undo record access in MVCC read. (MDEV-31355)

  • With InnoDB storage engine, when executing a statement that rebuilds an InnoDB table (such as TRUNCATE TABLE, OPTIMIZE TABLE, ALTER TABLE, or as part of executing innodb_undo_log_truncate=ON), the server can hang. (MDEV-31442)

  • With InnoDB storage engine, when a BINARY(0) or VARBINARY(0) column in an InnoDB table is indexed, the server can crash. (MDEV-19216)

  • Setting session_track_system_variables globally to an invalid value can cause the server to crash. (MDEV-25237)

  • Replication from an older MariaDB Server version to a newer MariaDB Server version can break, and the server may crash. (MENT-1935)

    • A mismatch in hash values was caused by use of different hash functions, causing rows in tables having explicit or implicit unique hash indexes to be treated as different rows between different versions of MariaDB Server even though the data in the rows was the same.

Can result in unexpected behavior

  • For a System Versioned table with non-versioned columns, if the initial INSERT includes a versioned column, an "on duplicate key update" for the non-versioned column generates a history record. (MDEV-23100)

  • With Spider storage engine, setting a system variable overrides the value set as a table parameter. (MDEV-31524)

  • The "pam_user_map" module does not correctly handle usernames or group names containing the "@" character. (MDEV-31336)

  • With some Unicode collations, JSON functions can return incorrect results. (MDEV-23187)

  • With sql_mode=ORACLE, information_schema.ROUTINES.ROUTINE_DEFINITION does not correctly contain the PACKAGE or PACKAGE BODY (MDEV-30662)

  • Stored routines with ROW variables set by subselects can produce incorrect results. (MDEV-31250)

  • Integer multiplication, DIV, MOD, or ROUND/TRUNCATE could return unexpected results when an argument is evaluated as -9223372036854775808 (MDEV-30932)

  • With Galera Cluster, when using a KILL QUERY statement on a Galera node that executes user transactions, GTID sequence drift can occur between writer and appliers nodes. At the same time, data consistency is maintained and wsrep_last_committed is in sync. (MDEV-31075)

  • With Galera Cluster, creating a TEMPORARY SEQUENCE can cause inconsistency. (MDEV-31335)

  • With Galera Cluster, the state of the cluster can only be retrieved from the primary component. (MDEV-21479)

  • information_schema.PARAMETERS can include outdated data when a stored routine is changed in one session while the stored routine is being used in a second session. (MDEV-31064)

  • Queries that use multiple RANK window functions can produce the wrong result. (MDEV-20010)

  • Queries that use the < "less than" operator to compare a string with a prefixed BLOB key produce the wrong result. (MDEV-31800)

  • Recursive CTE execution is interrupted without errors or warnings when max_recursive_iterations is reached. (MDEV-31214)

    • Starting with this release, a warning occurs when max_recursive_iterations is reached:

      Warning 1931 Query execution was interrupted. The query exceeded max_recursive_iterations = 1000. The query result may be incomplete.

  • On Microsoft Windows, when lower_case_table_names=2, SHOW TABLES does not work properly. (MDEV-30765)

  • When the system is busy, STOP REPLICA can take a long time. (MDEV-13915)

  • With parallel replication, Seconds_Behind_Master can show a wrong value. (MDEV-30619)

  • With replication for System Versioned tables, having a parent table and a child table WITH SYSTEM VERSIONING where child table has a Foreign Key CASCADE generates orphan rows on replica. (MDEV-31313)

  • With replication, when in read-only mode, XA COMMIT/XA ROLLBACK on the replica fails to return an error. (MDEV-30978)

  • mariadb-dump --force can stop with the error Couldn't execute 'SHOW CREATE FUNCTION `object`': even though --force should cause the error to be ignored. (MDEV-31092)

  • A query using DISTINCT, LIMIT with offset, and subselect to join tables produces an unexpected result. (MDEV-28285)

  • With the ColumnStore storage engine and Federated storage engine, ANALYZE can return the incorrect value 0 for r_rows (MDEV-29284)

  • With InnoDB storage engine, SET GLOBAL innodb_undo_log_truncate=ON does not free space when no undo logs exist. (MDEV-31382)

  • With InnoDB storage engine, freed data pages are not always scrubbed. (MDEV-31253)

  • With InnoDB storage engine, when setting innodb_open_files to a number lower than the number of tables used, an error can be shown in the server log and the operation will be aborted. (MDEV-31256)

  • ALTER TABLE .. MODIFY COLUMN can break foreign key constraints and lead to unrestorable dumps. (MDEV-31086)

  • With InnoDB storage engine, three concurrent DELETE by a UNIQUE key can cause an unexpected deadlock. (MDEV-10962)

  • With InnoDB storage engine, innochecksum fails with Floating point exception error. (MDEV-31641)

  • With HashiCorp key management plugin, possible memory leaks. (MENT-1874)

  • Using functions MAX() or MIN() with functions ROUND(time), CEILING(time), or FLOOR(time) as an argument can return wrong results. (MDEV-23838)

  • For transaction precise System Versioned tables, UPDATE can return an unexpected error: ERROR 1761 (23000): Foreign key constraint for table 'xxx', record 'yyy' would lead to a duplicate entry in table 'xxx', key 'PRIMARY' (MDEV-25644)

  • Assertion const_item_cache == true failed in Item_func::fix_fields when a flow control statement (such as IF()) was used in a generated column. (MDEV-31319)

  • Creating a table with a foreign key (with a cascade action) defined on a base column of a virtual column is not rejected. (MDEV-18114, MDEV-31322)

    • Starting with this release, it is no longer possible to create STORED generated columns and CHECK constraints when values of the affected columns can be changed by foreign key constraint actions, such as SET NULL or ON UPDATE CASCADE.

      • Starting with this release, this results in an error like: ERROR 1901 (HY000): Function or expression 'f_id' cannot be used in the GENERATED ALWAYS AS clause of 'v_id'

    • Starting with this release, for existing tables with STORED generated columns, SET NULL and ON UPDATE CASCADE are ignored.

Interface Changes

Spider Storage Engine

Platforms

In alignment to the enterprise lifecycle, MariaDB Enterprise Server 10.5.22-16 is provided for:

  • CentOS 7 (x86_64)

  • Debian 10 (x86_64, ARM64)

  • Debian 11 (x86_64, ARM64)

  • Microsoft Windows (x86_64) (MariaDB Enterprise Cluster excluded)

  • Red Hat Enterprise Linux 7 (x86_64)

  • Red Hat Enterprise Linux 8 (x86_64, ARM64)

  • Red Hat Enterprise Linux 9 (x86_64, ARM64)

  • Rocky Linux 8 (x86_64, ARM64)

  • Rocky Linux 9 (x86_64, ARM64)

  • SUSE Linux Enterprise Server 12 (x86_64)

  • SUSE Linux Enterprise Server 15 (x86_64, ARM64)

  • Ubuntu 20.04 (x86_64, ARM64)

Some components of MariaDB Enterprise Server might not support all platforms. For additional information, see "MariaDB Corporation Engineering Policies".