>> Hi all,
>>
>> I just came across following scenario:
>>  - using resource group DRBD - Delay - Xen
>>  - resources are active on node A that should become active
>>  - Start CRM
>>  - CRM probes resources, detects Delay as Stopped
>>  - CRM wants to start it, because of dependancy meaning Stop VM, Start
>> Delay, Start VM
>>
>> I think restart of VM is not needed, it is caused due to fact that
>> Delay RA is not stateful. Would you agree if it is modified to become
>> stateful? If yes, I could supply a patch.
>
>
> If you have to resort to use the "Delay" resource agent,
> that almost always means there is a bug somewhere else,
> either in some resource agent, or interaction, or
> in the configuration concept or dependencies.
>
> Please don't mask bugs.

My apologies, I hurried a bit.
I checked the details now and can confirm that OCF Delay RA _is_
stateful. The thing that happened to me was caused by fact that
Heartbeat cleans up $RSC_TMPDIR on every start. This directory is used
for state information handling by default. So it was my configuration
failure because I relied on it.

However, it uncovered that behavior of such resources could differ
whether or not CRM uses Heartbeat or AIS. At least looking at
Pacemaker code (1.0.7), I only found creation of this directory but no
clean up. I would not say it is a bug though, just a Heartbeat legacy
that users should be aware of.

Thanks,
Tino

>
> --
> : Lars Ellenberg
> : LINBIT | Your Way to High Availability
> : DRBD/HA support and consulting http://www.linbit.com
>
> DRBD? and LINBIT? are registered trademarks of LINBIT, Austria.
>
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to