Codership is pleased to announce a new Generally Available (GA) release of the multi-master Galera Cluster for MySQL 8.0, consisting of MySQL-wsrep 8.0.37-26.19 (release notes, download), with Galera replication library 4.19 (release notes, download) implementing wsrep API version 26. This release incorporates all changes to MySQL 8.0.37, adding a synchronous option for your MySQL High Availability solutions.
In the Galera replication library 4.19, some notable changes include the phasing out of the WSREP_KEY_SHARED key type since it had insufficiently strict certification rules that may lead to certification test passes when it should fail. It now only exists to support backwards compatibility, and new writesets are constructed using WSREP_KEY_REFERENCE. Network level node isolation helps shut down all communication with the rest of the cluster to achieve a fail-silent crash. The GCS level protocol has been bumped up, as error voting for joiner in the JOINED state was broken when group-wide commit cut (implied SUCCESS vote) was not taken into account when processing the error vote requests from the JOINED node. In addition, a memory leak in the commit cut was fixed.
In MySQL 8.0.37, we found that running GRANT with wsrep_forced_binlog_format=ROW turned on, produced an unrecoverable binary log. Now when you turn on wsrep_forced_binlog_format it is only effective for DML operations. In addition, it is also deprecated because binlog_format has been deprecated since MySQL 8.0.34, as the only possible logging format is ROW, making this option redundant. Table names with special characters have wider support in Galera Cluster, and binlog coordinated wsrep recovery has improved too.
There is the occasional stuck on starting commit state, and to fix this we reorganised the BF abort codepaths so that mutex locking cycles don’t happen when accessing the victim THD from BF THD. This was fixed in MDEV-29293 and backported. Small change, but WSREP_SST_OPT_PORT unset variable check fixed. If you read the GRA log files with mysqlbinlog, there have been occasions where it is malformed, due to a missing Format_description_log_event, now fixed. There was a race condition in wsrep InnoDB watchdog. A fix applied for when an unexpected server abort after MDL conflict between RSU and local command. MDL conflict reporting is now more verbose, as all lock types are printed. Fix applying failures due to MDL deadlocks by making sure that appliers get higher priority compared to locally executing DDLs in MDL deadlock resolving. Lastly, a warning has been improved for when you see "BF applier failed to open_and_lock_tables".
Please download the latest software and update your Galera Clusters! We continue to provide repositories for popular Linux distributions, and we encourage you to use them. Contact us more more information about what Galera Cluster Enterprise Edition can do for you.