MQSeries.net :: View topic - Getting Pacemaker clustering to work with MQ (not for RDMQ)

jamesb

Hi all,
Anyone using Pacemaker as the clustering software for active-active 2 node MQ clusters? I'm referring to basic 2 node clusters to control failover between nodes. Note this is not an RDQM set up, just HA.

I've previously used Veritas VCS for this, but my hands are tied on company standards. We are moving from MQ 8/Redhat 6/Redhat clustering to MQ 9.1/Redhat 7/Pacemaker clustering. If so:
1. Did you use a systemd unit file to start/stop MQ or an OCF script to stop/start MQ within the cluster?
2. If using the systemd unit files called by Pacemaker to start/stop MQ, did you use any specific options set within the unit file?
3. Did you set resource stickiness to stop MQ failing over unneccesarily?
4. We are experiencing some issues with MQ being failed over inadvertently due to occasional high CPU spikes which seems to be related to the fencing device. Is it possible to prevent failover due to high CPU? We are going to put in a ticket with Redhat as it seems to be a problem other (non-MQ) users have experienced.
Thanks, James.