MariaDB 10.10 is a current short-term support development series of MariaDB. It is an evolution of MariaDB 10.9 with several entirely new features.
For an overview of MariaDB 10.10 see the What is MariaDB 10.10? page.
Thanks, and enjoy MariaDB!
Notable Items
InnoDB
- InnoDB corruption due to lack of file locking (MDEV-28495)
- FULLTEXT search with apostrophe, and mandatory words (MDEV-20797)
- ALTER TABLE IMPORT TABLESPACE corrupts an encrypted table (MDEV-28779)
- ALTER TABLE wrong-result fix (MDEV-26294)
- Crash recovery fixes (MDEV-28668, MDEV-28731)
- DDL crash recovery fixes (MDEV-28752, MDEV-28802, MDEV-28864, MDEV-28870, MDEV-28923, MDEV-28977)
- Avoid crashes on corrupted data (MDEV-13542, MDEV-18519, MDEV-21098, MDEV-22388, MDEV-28457, MDEV-28950)
- Bulk load bug fixes (MDEV-28242, MDEV-28679)
- Performance fixes (MDEV-28708, MDEV-28766)
- Remove innodb_version (MDEV-28554)
- Deprecate and ignore the parameter innodb_prefix_index_cluster_optimization (MDEV-28540)
- Some InnoDB counters are duplicating generic SHOW STATUS (MDEV-28539)
- Useless output in SHOW ENGINE INNODB STATUS (MDEV-28542)
Replication
- ER_SLAVE_INCIDENT error is specified now on slave to be seen with SHOW-SLAVE-STATUS (MDEV-21087)
- INCIDENT_EVENT is no longer binlogged when a being logged transaction can be safely rolledback (MDEV-21443)
- sequences related row-format events are made to correspond to binlog_row_image (MDEV-28487)
- Possible reason of FLUSH BINARY LOGS hang is eliminated (MDEV-28948)
- Fix out-of-order gtid error in the circular semisync setup (MDEV-28609)
- Added global.slave_max_statement_time option for SQL thread to limit maximum execution time per query replicated (MDEV-27161)
- MASTER_USE_GTID=Current_Pos of CHANGE MASTER is deprecated to favor new MASTER_DEMOTE_TO_SLAVE option (MDEV-20122)
- MASTER_USE_GTID defaults of CHANGE MASTER TO and RESET SLAVE are changed to be compatible with GTID-based replication (MDEV-19801)
Galera
- Possible to write/update with read_only=ON and not a SUPER privilege (MDEV-28546)
- Node crashes with Transport endpoint is not connected mysqld got signal 6 (MDEV-25068)
- Galera4 not able to report proper wsrep_incoming_addresses (MDEV-20627)
- Galera should replicate nextval()-related changes in sequences with INCREMENT <> 0, at least NOCACHE ones with engine=InnoDB (MDEV-27862)
- Add support for OpenSSL 3.0 in Galera (MDEV-25949)
- Implement a method to add IPs to allowlist for Galera Cluster node addresses that can make SST/IST requests (MDEV-27246)
Optimizer
- Server crash in JOIN_CACHE::free or in copy_fields (MDEV-23809)
- Queries that use DISTINCT and an always-constant function like COLLATION(aggegate_func(…)) could cause a server crash. Note that COLLATION() is a special function – its value is constant even if its argument is not costant.
- Crash when using ANY predicand with redundant subquery in GROUP BY clause (MDEV-29139)
- A query with a subuquery in this form could cause a crash:
... ANY (SELECT ... GROUP BY (SELECT redundant_subselect_here)) ...
- MariaDB Server SEGV on INSERT .. SELECT (MDEV-26427)
- Certain queries in form “INSERT … SELECT with_aggregate_or_window_func” could cause a crash.
- restore_prev_nj_state() doesn’t update cur_sj_inner_tables correctly (MDEV-28749)
- Subquery semi-join optimization could miss LooseScan or FirstMatch strategies for certain queries.
- Optimizer uses all partitions after upgrade to 10.3 (MDEV-28246)
- For multi-table UPDATE or DELETE queries, the optimizer failed to apply Partition Pruning optimization for the table that is updated or deleted from.
- Range optimizer regression for key IN (const, ….) (MDEV-25020)
- The issue can be observed on MariaDB 10.5.9 and later versions which have the fix for MDEV-9750. That fix introduceds optimizer_max_sel_arg_weight.
- If one sets optimizer_max_sel_arg_weight to a very high value or zero (which means “unlimited”) and runs queries that produce heavy-weight graphs, they can observe a performance slowdown, e.g.:
table.keyXpartY [NOT] IN ( ... )
- Wrong result with table elimination combined with not_null_range_scan (MDEV-28858)
- If one runs with optimizer_switch=’not_null_range_scan=on’ (which is not enabled by default), a query that does a join and has const tables could produce a wrong result.
- Assertion `tmp >= 0′ failed in best_access_path (MDEV-28882)
- If one uses histogram_type=JSON_HB, has collected a histogram of that type and runs a query that selects a very narrow range near histogram end, they can hit an assertion in the optimizer due to rounding errors in the histogram causing negative selectivity.
General
- Crash in JSON_EXTRACT (MDEV-29188)
- ALTER TABLE ALGORITHM=NOCOPY does not work after upgrade (MDEV-28727)
- Server crash upon CREATE VIEW with unknown column in ON condition (MDEV-29088)
- password_reuse_check plugin mixes username and password (MDEV-28838)
- As per the MariaDB Deprecation Policy, this will be the last release of MariaDB 10.10 for Debian 10 “Buster” for ppc64el
Changelog
For a complete list of changes made in MariaDB 10.10.1, with links to detailed information on each push, see the changelog.
Do not use non-stable (non-GA) releases in production!