Percona XtraDB Cluster - HA scalable solution for MySQL

Comment 6 for bug 931713

Alex Yurchenko (ayurchen) wrote :

Hi Patrick,

I'm not sure what you're trying to achieve there changing wsrep_cluster_address randomly without regards to node state. Changing wsrep_cluster_address is intended for rational use, it is a human driven event, not a random disaster, and from your report it looks like you were testing cluster resilience to human intervention. :)

Now, wsrep_cluster_address value in itself is of little interest to us, it is a just string. Far more interesting is what this changing it causes in Galera. Wsrep status variables and error log would tell a lot about what is _really_ going on. You should always check cluster and node state before giving it administration command.

I'm not denying that there could be a bug, however 2.1 will see a lot of changes in this department, so I'd suggest to return to this after 2.1 release.