[Engine-devel] SSH Soft Fencing
Yaniv Bronheim
ybronhei at redhat.com
Thu Jul 18 06:48:17 UTC 2013
----- Original Message -----
> From: "Itamar Heim" <iheim at redhat.com>
> To: "Martin Perina" <mperina at redhat.com>
> Cc: engine-devel at ovirt.org
> Sent: Tuesday, July 16, 2013 3:27:18 PM
> Subject: Re: [Engine-devel] SSH Soft Fencing
>
> On 07/03/2013 02:55 PM, Martin Perina wrote:
> > Let's summarize again, SSH Soft Fencing patches has been merged yesterday
> > with following functionality:
> >
> > 1) For hosts with power management configured, SSH Soft Fencing is the 1st
> > fencing stage. If it doesn't help, real fencing will be executed.
> >
> > 2) For hosts without power management configured, SSH Soft Fencing is the
> > only
> > fencing stage. If it doesn't help, host will become non responsive.
> >
> > 3) SSH Soft Fencing is enabled by default, there's no configuration option
> > to disable it
> >
> > 4) SshSoftFencingCommand option is used to define what command is executed
> > during SSH Soft Fencing. It can only be changed manually in database.
> >
> > The whole fencing process in oVirt 3.3 is decribed at
> >
> > http://www.ovirt.org/Automatic_Fencing#Automatic_Fencing_in_oVirt_3.3
>
> I assume ssh soft fencing isn't considered "fencing"? i.e., i assume it
> does not release resource occupied by the host (spm, running vms)?
>
it should.
the correct way to release resources when killing vdsm, as it implemented now, is to send SIGTERM to vdsm process, as sanlock does. but it might take time (more than 40 sec)
currently we just kill and start over the process with the service tool. SIGTERM calls to prepareForShutdown, it might be necessary if after the restart the engine initiates failover.
"stop" verb does send SIGTERM before SIGKILL to vdsm, we just need verify that it waits enough time before sending the KILL and vdsm initiated the prepareForShutdown scope.
> Thanks,
> Itamar
>
> _______________________________________________
> Engine-devel mailing list
> Engine-devel at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/engine-devel
>
More information about the Engine-devel
mailing list