On 2009-11-16T18:58:12, Dejan Muhamedagic <[email protected]> wrote:
> To have this handled by CRM, crmd would have to cancel the
> currently running stonith action, i.e. send the appropriate
> message to stonithd.
Sure. This also happens if the node eventually reboots and rejoins
cleanly, too, anyway.
> If handled by stonithd, it would have to cancel the running stonith
> action and send the OK status to crmd. It basically boils down to the
> same, but the former needs extra support in crmd. Don't know how much
> effort that would take and if it would make the code more complex.
This might work too, yes - basically faking a stonithd "success".
However, how do you send this if there's no currently pending STONITH,
because stonith has just timed out and been handed back to the
transitioner?
Sending the fake ack to crmd might be the right way.
> BTW, this would obviate the need for this patch, but not for the
> meatware plugin, since the CRM would complain if there are no
> stonith resources in the configuration with fencing enabled.
What I really dislike about the meatware plugin is that it needs to be
run on a specific node. It'd be much better if it could run on any
node.
And there should always be some stonith resource defined. If the manual
override existed, they could simply define external/ssh even (no need
for meatware), which would at least allow error recovery for, say, stop
failures.
Regards,
Lars
--
Architect Storage/HA, OPS Engineering, Novell, Inc.
SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde
_______________________________________________________
Linux-HA-Dev: [email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/
