wsrep_cluster_size variable resetting to zero

XtraDB Cluster Version 5.5.29

Odd behaviour that cannot explain.

The wsrep_cluster_size variable we use to monitor (Nagios) the cluster as it is one of the status variables… obviously if the number of members drops we could have a problem or split brain so we need to know.

However we are seeing this value drop to zero, nothing in the errors logs on any node primary or otherwise explains it.

It happens on the primary, the numbers are fine on the other members.

You can get it to reset by restarting one of the members (not the primary). On a secondardy restart mysql and the number chnages say from 0 to n-1 and when the secondary is back goes to n where n is the number of members in the cluster.

Nothing seems to be wrong but as this creates alerts it is a concern.

Anyone else seeing this?

Dom

Can you share the MySQL error log from the “primary” when this happens? This should not be happening on any of the remaining nodes, do you have identical hardware and configuration for each node? What does your load look like?

Thanks… I am a contractor and my contract ended, I only just noticed your reply … the client will have to figure this out for themselves now.

Dom