Using Corosync Pacemaker solution inconjunction to Percona xtraDB Cluster

Hi ,

We are currently hosting a 3-node Percona xtradb cluster for one of the clients and are using the corosync pacemaker software - which basically uses a floating IP to point to the primary node in the cluster. This was implemented to prevent the application failing to connect to the DB when a failover occurs. So that the application need not re-configure the connection string every time.
Recently we had an incident where the MySQL service failed on one of the nodes and the floating IP address did not automatically failover to the new primary. It was then I learnt that corosync pacemaker cannot detect the MySQL process failure but only if the entire node goes down it automatically fails over to the new primary.

My question: Is there a way we can detect with corosync pacemaker if the MySQL process fails and it will automatically failover to the new primary? Has anyone used this solution before and if so what was the process followed. Some insight on this topic will be really helpful.


This depents on your Pacemaker agent to detect what happens to the active master. If you are using the default agent, I think it has limited monitoring options and cannot handle PXC/Galera. I’m not sure if anyone has a custom agent for this.

If you want some form of VIP handling, perhaps HAProxy route would be a more suitable solution for now?