We encountered an issue about the fencing with Pacemaker (1.1.10-*) and corosync on our clusters.<br />
Indeed, when we suddenly stop any resource via kill -9 (MySQL or IPaddr for example), the monitor detects a problem (expected behaviour) but it cannot execute the fencing. <br />
All of these resources are configured with the “fence” option for the “on-fail” parameter.<br />
<br />
Please note this issue is not encountered when we kill corosync. In this case, the fencing is well done.<br />
<br />
To fix it, we had to downgrade the Pacemaker package (1.1.8-7.el6).<br />
<br />
Don’t hesitate to ask me if you need more information.
↧