On Fri, Feb 27, 2009 at 22:32, Andreas Mock <[email protected]> wrote: >> -----Ursprüngliche Nachricht----- >> Von: "Andrew Beekhof" <[email protected]> >> Gesendet: 16.02.09 11:23:53 >> An: General Linux-HA mailing list <[email protected]> >> Betreff: Re: [Linux-HA] Generell STONITH resource configuration question > > Hi all, > >> > >> > Just to get it right, I'm using IPMI as STONITH device. I have to set up >> > two resources in a 2-Node cluster and add constraints in order to force >> > each resource to run on a separate nodes. >> > >> > Is this approach correct? Anything else I need to worry about when >> > creating the constraints / resources? >> >> You might be able to run it as a clone which would eliminate the need >> for any constraints. > > I want to emphasis one issue. Some of the stonith devices (not stonith > plugins) > are not "cloneable". I don't know if it's true for IPMI but there are stonith > devices > which do allow only ONE connection at a time. With a "cloned" config you can > NOT > define that the status call to the stonith plugins shall be done in a > controlled > sequence. That means, IMHO, if you use a "cloned" config it can happen that > both (2-node cluster) sides try to do the status operation at the same time. > One of them can be rejected (=> bad status).
Another option for such devices might be to use a Master/Slave and only have the master do monitoring. I wonder if the lrm hooks for stonith can handle this. > We discussed this once and came to the conclusion that a stonith plugin > developer should document if this plugin is "cloneable" reliably. > > Correct me if something changed the last months in the stonith subsystem > which does falsify my "theory". > > Best regards > 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
