Bugs fixed¶

XtraBackup SST wouldn’t stop when MySQL was SIGKILLed. This would prevent MySQL to initiate a new transfer as port 4444 was already utilized. Bug fixed #1380697.

wsrep_sst_xtrabackup-v2 script was causing innobackupex to print a false positive stack trace into the log. Bug fixed #1407599.

MyISAM DDL (CREATE/DROP) isn’t replicated any more when wsrep_replicate_myisam is OFF. Note, for older nodes in the cluster, wsrep-replicate-myisam should work since the TOI decision (for MyISAM DDL) is done on origin node. Mixing of non-MyISAM and MyISAM tables in the same DDL statement is not recommended with wsrep_replicate_myisam OFF since if any table in list is MyISAM, the whole DDL statement is not put under TOI (total order isolation). Bug fixed #1402338.

gcache.mem_size has been deprecated. A warning will now be generated if the variable has value different than 0. Bug fixed #1392408.

stderr of SST/Innobackupex is logged to syslog with appropriate tags if sst-syslog is in [sst] or [mysqld_safe] has syslog in my.cnf. This can be overridden by setting the sst-syslog to -1 in [sst]. Bug fixed #1399134.

clustercheck can now check if the node is PRIMARY or not, to allow for synced nodes which go out of PRIMARY not to take any writes/reads. Bug fixed #1403566.

SST will now fail early if the xtrabackup_checkpoints is missing on the joiner side. Bug fixed #1405985.

socat utility was not properly terminated after a timeout. Bug fixed #1409710.

When started (without bootstrap), the node would hang if it couldn’t find a primary node. Bug fixed #1413258.

10 seconds timeout in Percona XtraBackup SST Configuration script was not enough for the joiner to delete existing files before it started the socat receiver on systems with big datadir. Bug fixed #1413879.

Non booststrap node could crash while attempting to perform table%cache operations with the BF applier failed to open_and_lock_tables warning. Bug fixed #1414635.

Percona XtraDB Cluster 5.6 would crash on ALTER TABLE / CREATE INDEX with Failing assertion: table->n_rec_locks == 0. Bug fixed #1282707.

Variable length arrays in WSREP code were causing debug builds to fail. Bug fixed #1409042.

Race condition between donor and joiner in Percona XtraBackup SST Configuration has been fixed. This caused XtraBackup SST to fail when joiner took longer to spawn the second listener for SST. Bug fixed #1405668.

Signal handling in mysqld has been fixed for SST processes. Bug fixed #1399175.

SST processes are now spawned with fork/exec instead of posix_spawn to allow for better cleanup of child processes in event of non-graceful termination (SIGKILL or a crash etc.). Bug fixed #1382797.

wsrep_local_cached_downto would underflow when the node on which it is queried had no writesets in gcache. Bug fixed #1262179.

A typo in wsrep_provider_options could cause an unhandled exception. Bug fixed #215.

Interrupted IST would result in HA_ERR_KEY_NOT_FOUND error in subsequent IST. Bug fixed #210.

garb logger was showing incorrect error message. Bug fixed #168.

Other bugs fixed: #1275814.

Known Issues¶

For those affected by crashes on donor during SST due to backup locks (#1401133), please add the following to your my.cnf configuration file:

[sst]
inno-backup-opts='--no-backup-locks'

option as a workaround to force FLUSH TABLES WITH READ LOCK (NOTE: This workaround will is available only if you’re using Percona XtraBackup 2.2.9 or newer.). Or as an alternative you can set your environment variable FORCE_FTWRL to 1.

Help us improve quality by reporting any bugs you encounter using our bug tracking system. As always, thanks for your continued support of Percona!