[ClusterLabs] STONITH interrupting pacemaker yum upgrades

[Ken Gaillot]

On 02/19/2016 09:50 AM, Richard Stevenson wrote:
> Hi,
> 
> I'm having trouble updating pacemaker on a small 3 node cluster. All nodes
> are running Centos 7, and I'm upgrading via a simple `yum upgrade`.
> Whenever I attempt to do this the node is fenced when yum attempts to clean
> up the old pacemaker package - restarted by STONITH in this case.
> 
> Updating pacemaker is a common requirement, is there a recommended process
> to upgrade pacemaker without taking out the node, or should I remove the
> node before upgrading HA components?
> 
> Thanks,
> Richard Stevenson

In most cases, you should stop the cluster on a node before upgrading it
("pcs cluster stop" on Fedora/RHEL/CentOS). The document below describes
various approaches to updates:

http://clusterlabs.org/doc/en-US/Pacemaker/1.1-pcs/html-single/Pacemaker_Explained/index.html#ap-upgrade

This reminds me of a new feature that I'll post about separately ...