Hi, On Thu, Oct 25, 2007 at 11:34:19AM +0200, matilda matilda wrote: > >>> Dejan Muhamedagic <[EMAIL PROTECTED]> 25.10.2007 11:17 >>> > >> > >> One per machine and resticted to run on the other...what have you thought? > >> ;-)) > > > >That should suffice. BTW, I don't think this (don't wanna kill > >meself) is a very serious issue, because a typical situation is > >that another node does a reset. > > Hi Dejan, > > I agree with you. The "problem" arises when I get a failure on the > monitor action of the stonith plugin, failcount and the corresponding > calculation forces the stonith resource to be moved to exactly that > node which has to be shot. My intent is to have a fallback situation > in which the stonith plugin can (!) be moved to the node which has to > be killed. > > As far as I understand it now: If this would happen, I'll see the > stonith resource moved to the other node, feel happy about the > nice failover mechanisms but the stonith agent will prevent to shoot > itself. In this case I can get rid of the whole failover mechanism.
Perhaps you have a good point here. I have to think about it more. However, I don't really understand why should a stonith resource be happy to run on host A but not on host B. It's talking to the same device. Is it that the monitor operation fails occasionaly? Perhaps that's what you should fix. If you can't, maybe set up this resource not to fail over immediately. Perhaps the cluster can also forget (somehow) about the failures which happened in the distant past. But, yes, basically in this situation a fail over makes no sense. > This is the motivation to discuss about it. Thanks, Dejan > Happy digging > Andreas Mock > > > _______________________________________________ > 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
