Hi Thanks but no, in older releases, the op monitoring failed leaded to "fence" as required by "on-fail=fence" . If you do not set on-fail=fence, you're right, Pacemaker restart the resource either on local node if migration-threshold>1 or on adjacent node if migration-threshold=0, but if on-fail=fence, the other node should fence the node immediately after getting the monitoring failed.
Alain De : Florian Crouzat <[email protected]> A : [email protected] Date : 12/11/2012 14:40 Objet : Re: [Linux-HA] Bug around on-fail on op monitor ? Envoyé par : [email protected] Le 12/11/2012 13:27, [email protected] a écrit : > Hi Andrew, > > I'm using pacemaker-1.1.7-6 > > On my resources I set on-fail=fence on the op monitor, and to test it, I > just > stop manually the resource out of Pacemaker to lead to a monitoring failed > in Pacemaker, > but with this release the resource is restarted on same node , and the > other node does > not even try to fence the local node ... > > Besides, it was working fine in older releases, meaning the node was well > fenced in this test case. > > Is it a known bug or regression ? Is there a patch file somewhere ? I'm not Andrew but I believe that before declaring the resource as failed, on an unsuccessful op.monitor() the cluster first tries to restart the resource, which in your case, will succeed. But I cannot find anything in the documentation to prove that, sadly. You'd rather use: crm resource failcount foo set X (or return a wrong exit code from the init-script, or make sure that the service cannot start anymore) -- Cheers, Florian Crouzat _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
