"wsrep_local_replays" greater than zero
I've been having this situation with two different Galera DB Clusters (let's say Cluster A and Cluster B) for two different customers, where the "wsrep_local_replays" shows "1" on one of the DB nodes on each DB Cluster.
I'd like to know:
1. Which is reason why this variable is now 1. 2. How can this be fixed. 3. As Cluster A and Cluster B are in Production, I'm worried that this may cause a problem in the long term... Would it cause it?
One fix that worked for me was restarting mysql service on the node with this "issue" in Cluster A, which thankfully joined the cluster with an IST.
However, I cannot do the same for Cluster B, since DBs are massive ( > 300GB). Even though gcache is big enough, that won't guarantee an IST and I cannot afford to do an SST. Also, the donor is the one having "wsrep_local_replays=1". The app doesn't write to the donor (so how did it happen?).
Cluster A: 10.0.19-MariaDB-wsrep-log MariaDB Server, wsrep_25.10.r4144
Cluster B: 5.5.41-MariaDB-wsrep-log MariaDB Server, wsrep_25.11.r4026
Please let me know if you need any more details.
Thanks in advance!
Answer Answered by Jan Lindström in this comment.
wsrep_local_replays tells you total number of transaction replays due to asymmetric lock granularity. It is just a status variable and its value naturally can be 1 if during uptime one transaction was replayed locally. There is no need for IST/SST in this case.