SST and async slaves

MySQL & MariaDB Percona XtraDB Cluster 5.x
1

Hi,

i’d like to know if i have to rebuild the async slaves from scratch (stop slave, create dump, load dump, …) of a pxc node (5.6.15-25.5-759.wheezy) in case the pxc node performs an SST (xtrabackup-v2), e.g. after a crash or long downtime ? As it seems that SST does not transfer the binlogs, the binlogs can’t be consistent with the rest (i.e. the state transferred by SST) after the SST has finished, because the state transfer probably includes intermediate changes, which are then missing in the binlogs.

Thanks for your help

2

If async master node indeed needs full SST for some reason, then it means it got out of sync with cluster probably, hence it may be no longer reliable source of data for async slave. The easiest solution would be using PXC 5.6 and GTID enabled binlogs, so when async master node has problems - you just switch to another node also having GTID binlogs enabled.
[URL=“Changing an async slave of a PXC cluster to a new Master using 5.6 and GTID”]http://www.percona.com/blog/2014/02/...sing-5-6-gtid/[/URL]