[ClusterLabs] Antw: [EXT] Re: VirtualDomain does not stop via "crm resource stop" - modify RA ?

[Lentes, Bernd]

----- On Oct 26, 2020, at 8:41 AM, Ulrich Windl Ulrich.Windl at rz.uni-regensburg.de wrote:

> "SIGKILL: Device or resource busy" is nonsense: kill does not wait; it either
> fails or succeeds.

yes and no. When you send a SIGKILL to a process which is in 'D' state, the signal can't be delivered, e.g.  because the domain is doing heavy IO.
But the signal is pending, and when the process wakes up and switches to 'S' or 'R' state, the signal is delivered and the process is killed.

That's exactly what i want to address. If you just check the rc from kill and you get s.th. -ne 0, you think it failed.
But it's possible your kill will work 20 seconds later. So, with a delay, SIGKILL would succeed. But the node the domain is running on is already fenced, with all its consequences.

And the "device or resource busy" message is exactly the one you get when you do a "virsh destroy" to a domain in 'D' state.

Bernd
Helmholtz Zentrum München

Helmholtz Zentrum Muenchen
Deutsches Forschungszentrum fuer Gesundheit und Umwelt (GmbH)
Ingolstaedter Landstr. 1
85764 Neuherberg
www.helmholtz-muenchen.de
Aufsichtsratsvorsitzende: MinDir.in Prof. Dr. Veronika von Messling
Geschaeftsfuehrung: Prof. Dr. med. Dr. h.c. Matthias Tschoep, Kerstin Guenther
Registergericht: Amtsgericht Muenchen HRB 6466
USt-IdNr: DE 129521671