Hi,

On Wed, Jun 04, 2008 at 05:01:39PM +0800, Xinwei Hu wrote:
> 2008/6/4 Alexander Hofmann <[EMAIL PROTECTED]>:
> > Andrew Beekhof wrote:
> >>
> >> On Jun 4, 2008, at 10:07 AM, Alexander Hofmann wrote:
> >>
> >>> Andrew Beekhof wrote:
> >>>>
> >>>> On Wed, Jun 4, 2008 at 8:18 AM, Alexander Hofmann
> >>>> <[EMAIL PROTECTED]> wrote:
> >>>>
> >>>>> Hello list,
> >>>>>
> >>>>> I have a problem with my Heartbeat and STONITH configuration:
> >>>>> When I disconnect the two private links both nodes are STONITHed.
> >>>>>
> >>>>
> >>>> Right - because neither side can see the other.
> >>>> And because the cluster only has two nodes, both sides also have
> >>>> quorum (which allows them to shoot the other).
> >>>>
> >>>> Two options:
> >>>> - add a third node (so only one side of the cluster will have quorum)
> >>>> - set the "stonith-action" option to "poweroff" (which would prevent
> >>>> the shot node from shooting the survivor when it returns)
> >>>>
> >>>>
> >>> Hi,
> >>>
> >>> thank you for your fast reply.
> >>>
> >>> stonith-action ist already set to "poweroff".
> >>
> >> Thats bizare.  They're configured to shoot each other right (not
> >> themselves)?
> >
> > Yes, they are.
> >
> >> The chances of both nodes firing off a STONITH op at _exactly_ the same
> >> time, repeatably... it boggles the mind.
> >>
> >>> I've just tried to set ilo_powerdown_method to "power" and as a result
> >>> _both_ nodes do a normal shutdown...
> >>>
> >>> Any ideas aside from a third node?
> >>
> >> Not without logs from both machines.
> >
> > I copied some log to the bottom of my first mail or do you need more output?
> For your specific issue,
> set ilo_powerdown_method to power.

According to the docs, it should actually be set to "button". The
"power" method (which is also the default) involves ACPI which
usually results in a regular shutdown.

Thanks,

Dejan
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to