[Linux-HA] CRM and STONITH questions
dejanmm at fastmail.fm
Thu Oct 25 03:21:12 MDT 2007
On Wed, Oct 24, 2007 at 06:40:17PM +0200, Andrew Beekhof wrote:
> On 10/24/07, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
> > Hi,
> > On Wed, Oct 24, 2007 at 02:26:05PM +0200, Andrew Beekhof wrote:
> > > On 10/24/07, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
> > > > Hi,
> > > >
> > > > On Mon, Oct 22, 2007 at 09:21:22PM +1000, Spam Filter wrote:
> > > > > >HAv2 tries to shoot the node which has to be. So, it's possible that it
> > > > > >shoots itself, as far as I know. But why are you concerned about that?
> > > > >
> > > > > Only concerned because I saw another stonith script protect itself from
> > > > > shooting itself and I just wanted to make sure that HA2 isn't just
> > > > > sending the complete list for who knows what reason.
> > > > > If it's definitely sending only the hostlist of nodes it definitely
> > > > > wants to stonith then fine.. If it's possible that HA2 may think it
> > > > > should stonith itself and the HA community thinks it's something that
> > > > > does happen then that's fine too, I'll remove my anti-foot shooting
> > > > > parts I've just added in.
> > > >
> > > > The stonithd (i.e. the program which runs the stonith show)
> > > > should prevent suicides. Please let us know if it doesn't.
> > >
> > > really? how come?
> > Beat me. That's what it does in the code which I recently
> > reviewed. With the exception of testing, i.e. if it's run with
> > '-t'.
> i think thats a mistake
> a 1-node cluster could conceivably want to shoot itself in case of a
> resource failure
It may want to, but I don't see how would that improve things.
Actually, this may be a good reason to prevent shooting oneself.
> but i could be mistaken
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> See also: http://linux-ha.org/ReportingProblems
More information about the Linux-HA