Percona XtraDB Cluster 5.6.25-25.12 is now availableHrvoje Matijakovic
Based on Percona Server 5.6.25-73.1 including all the bug fixes in it, Galera Replicator 3.12, and on Codership wsrep API 25.11, Percona XtraDB Cluster 5.6.25-25.12 is now the current General Availability release. All of Percona’s software is open-source and free, and all the details of the release can be found in the 5.6.25-25.12 milestone at Launchpad.
- Percona XtraDB Cluster has implemented Support for PROXY protocol. The implementation is based on a patch developed by Thierry Fournier.
- MTR coverage has been added to all tests in galera suite. Many bugs associated with mtr tests have been fixed.
- A new variable
gcache.keep_pages_count, analogous to
gcache.keep_pages_size, has been added. The variable limits the number of overflow pages rather than the total memory occupied by all overflow pages. Whenever either of the variables are updated at runtime to a non-zero value, cleanup is called on excess overflow pages to delete them. This feature also fixes the bugs with integer overflow in the gcache module.
- Updates have been made to wsrep code to ensure greater concordance with binary log and GTID so that failover of async slaves, among nodes of the cluster is seamless and consistent. To ensure this in #1421360, all
FLUSH BINARY LOGand
FLUSH LOGS, and read lock-based flush such as
FLUSH TABLES WITH READ LOCKand
FLUSH TABLES FOR EXPORT),
ANALYZE TABLE, Percona Server-specific flush statements for user statistics and page tracking bitmaps are executed under Total Order Isolation (TOI) so that they are replicated to other nodes in the cluster when they are written to binary log.
- Percona XtraDB Cluster has temporarily disabled savepoints in triggers and stored functions. The reason is that even having fixed bug #1438990 and bug #1464468 we have found more cases where savepoints in triggers break binary logging and replication, resulting in server crashes and broken slaves. This feature will be disabled until the above issues are properly resolved.
SHOW STATUS LIKE ... and SHOW STATUSwere taking time proportional to size of gcache.size. Bug fixed #1462674.
- When disk space would get filled with
gcache.pagefiles, Galera would crash when the next page file was created. Bug fixed #1488535.
- XtraBackup SST didn’t clobber
backup-my.cnfwhich caused SST to fail. Bug fixed #1431101.
- Error from
process::waitwas not checked in joiner thread leading to joiner starting erroneously even when SST had failed. Bug fixed #1402166.
- Due to an regression introduced in Percona XtraDB Cluster 5.6.24-25.11, update of the wsrep_cluster_address variable, following the update of
wsrep_provider_optionsvariable would cause the server to deadlock.
- mysqldump SST could stall due to a regression in desync mutex introduced in Percona XtraDB Cluster 5.6.24-25.11 by fixing the bug #1288528.
mysql_tzinfo_to_sqlsets wsrep_replicate_myisam variable at session scope so that statements are replicated correctly.
Percona-XtraDB-Cluster-devel-56package was not included in the
Percona-XtraDB-Cluster-full-56metapackage on CentOS due to a conflict with upstream mysql package.
service mysql startand then
service mysql@boostrap startafterwards would cause server shutdown.
FLUSHcommands will ensure it is not written to binary log (as in mysql async replication) and not replicated in wsrep.
FLUSH TABLES WITH READ LOCKfailure (with non-existent tables) didn’t resume the galera provider, causing deadlock.
- Percona XtraDB Cluster will not blocking DDL statements on tables which are used with
... FOR EXPORTor
... WITH READ LOCK, it will give return an error message about read lock.
- Fixed the update of
wsrep_slave_threadsvariable regarding the default value assignment, invalid value truncation, and error issued while threads are still being closed.
- The mysql client in Percona XtraDB Cluster has been built with system
- Bugs in 32-bit galera associated with
available_storageand integer overflow after multiplication in offset calculation have been fixed.
- Galera 3 was failing to build on all non-intel platforms. Architecture specific
CCFLAGShave been removed and provision for inheriting
LDFLAGShave been added to SConstruct.
- Non-global read locks such as FLUSH TABLES WITH READ LOCK and FLUSH TABLES FOR EXPORT paused galera provider but didn’t block commit globally which
wsrep_to_isolation_begin(for DDL) was made aware of.
Help us improve our software quality by reporting any bugs you encounter using our bug tracking system. As always, thanks for your continued support of Percona!