From: "Michael Pasternak" <mpastern(a)redhat.com>
To: "Laszlo Hornyak" <lhornyak(a)redhat.com>
Cc: "engine-devel" <engine-devel(a)ovirt.org>
Sent: Sunday, January 27, 2013 9:27:54 AM
Subject: Re: [Engine-devel] Watchdog support Feature
On 01/24/2013 04:19 PM, Laszlo Hornyak wrote:
> Hi,
>
> Watchdog support in engine will add watchdog to the ovirt UI and
> REST API. Watchdog cards will be mainly used in HA vm's.
>
>
http://www.ovirt.org/Features/Watchdog_engine_support
>
> Your feedback is welcome!
>
> Thank you,
> Laszlo
> _______________________________________________
> Engine-devel mailing list
> Engine-devel(a)ovirt.org
>
http://lists.ovirt.org/mailman/listinfo/engine-devel
Definitely nice feature, but i have a question for [1],
won't it cause slow responsive guests (on highly loaded hosts)
to constantly reboot?
i.e it's understood that 'reboot' is a goal of this feature,
but endless reboot of pinned-to-host guests for instance
won't do any good right?,
my point is: should watchdog 'action' have extra logic
for guest-policy like placement for instance?
[1] <watchdog model="i6300esb" action="reset"/>
--
Michael Pasternak
RedHat, ENG-Virtualization R&D
Michael, good question.
The device itself will not be added by default,
and there are a few other actions which can be used (default action is none).
So this will be in use only by someone who expects it to reboot.
Additionally the watchdog is configurable to fire according to the internal
drivers settings, so softdog may be configured to trigger the watchdog by
several indicators and not only cpu load / time.