Hi, On Wed, Mar 12, 2008 at 04:05:02PM +0100, Andreas Mock wrote: > > -----Urspr?ngliche Nachricht----- > > Von: [EMAIL PROTECTED] > > [mailto:[EMAIL PROTECTED] Im Auftrag von > > Bryan Manzeck > > Gesendet: Dienstag, 11. M?rz 2008 16:58 > > An: linux-ha@lists.linux-ha.org > > Betreff: [Linux-HA] Node fences itself? > > Hi Bryan, > > > > > Hello, > > I have STONITH setup with HP iLO cards and I am wondering if > > the clone resource should run on itself? What I am trying to > > ask is if a node should be able to fence itself? Is there a > > best practice for this? > > Ha...that's a good question. We had some discussion about that > on the list. I really don't know the current status. So be careful > what I say: > > a) Dejan says that there is code in the stonithd to prevent a node > to shoot itself. BUT: I had a test case where exactly this happened. > I don't know what is really true, now.
stonithd does prevent nodes from shooting themselves, but not consistently. There's a case in which a stonith request slips through and the node does shoot itself. BTW, this case applies to a single node cluster, so the suicide plugin would work there. > b) What I do: I have single primitive stonith resources for every > node I want to shoot. Every primitiv has a location constraint giving > the node which has to be shot by exactly this stonith resource a > big penalty, but NOT -INFINITY. > > c) Why the penalty: > 1) I want to run the stonith resource on any node BUT the node to stonith. > 2) In a two node scenario. As soon as one node dies the stonith resource > is moved to the last goo node. As soon as self stonithing is working, this > should be a valid layout. Self stonithing is not going to work like that, simply because one can't ensure that the outcome of the operation is sent to the requester. > d) In a more-than-2-node-cluster you should be able to use clones with > nodecount-1 clones and a contraint forbidding to run on the target node. > > Dejan: What do you think? What's the current status? This is the current plan: http://developerbugs.linux-foundation.org/show_bug.cgi?id=1853 Comments are welcome, of course. Cheers, Dejan > Best regards > Andreas Mock > > _______________________________________________ > Linux-HA mailing list > Linux-HA@lists.linux-ha.org > http://lists.linux-ha.org/mailman/listinfo/linux-ha > See also: http://linux-ha.org/ReportingProblems _______________________________________________ Linux-HA mailing list Linux-HA@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
