We could add a dom0_op to prod the 'virtual watchdog', and then hook
that in to Xen's current NMI watchdog routine as an extra liveness check.
(we should enable Xen's watchdog by default anyhow, if we don't
already).
We’d have to think further about how we wanted to do 'hot failover' of
dom0's. Should the slave and master dom0 monitor each other in user space,
or would we want to have Xen involved. Probably the former.
Ian
> -----Original Message-----
> From: xen-devel-admin@lists.sourceforge.net
> [mailto:xen-devel-admin@lists.sourceforge.net] On Behalf Of
> James Harper
> Sent: 01 February 2005 11:29
> To: xen-devel@lists.sourceforge.net
> Subject: [Xen-devel] watchdog
>
>
> Some of the testing i'm doing with iscsi is causing me to
> hang dom0 (my fault, but a pain nonetheless). Has anyone had
> any thoughts about a watchdog interface to xen? It should be
> able to have the same interface as the existing watchdog
> stuff, except that it would be xen listening in the
> background, and xen rebooting the machine if dom0 went silent.
>
> James
> NHuryjj
> jzy䖭ƶ8IzrjÞ{zjzy~z
>
NHS^隊X'uץ%Hb'rدybW變^.ǝj֛jǬ
kjkަ&zئy+!h(^ƚ+8%Iz 塧Zrrjm"ץ+{((yWzw^fj)b bޝׯzYb,y+m+-.ǟ+-bا~zw^