[Pacemaker] Unexpected resource restarts after putting a node in standby mode

[Florian Haas]

On 07/10/2009 01:33 PM, daniel peess wrote:
> hi florian,
> 
> On Mon, Jul 06, 2009 at 08:00:38AM +0200, Florian Haas wrote:
>> order o_drbd_before_iscsivg01 inf: ms_drbd_iscsivg01:promote
>> rg_iscsivg01:start
>> order o_drbd_before_iscsivg02 inf: ms_drbd_iscsivg02:promote
>> rg_iscsivg02:start
> 
> before it promotes drbd to master somewhere else it stops all depending 
> resources, as order constraints are by default symmetric.

Nack. If you reread the ptest output excerpt in my original post, you'll
notice that there is no resource transition whatsoever that prompts a
promotion of the DRBD resource that the restarted resource group is
constrained to. There's two drbd resources there, and two target/LU
resource groups. One DRBD resource gets promoted, and one resource group
gets moved. That's fine. But there's another DRBD resource that remains
untouched, and _its_ resource group gets needlessly restarted. That's my
point.

>> rg_iscsivg02 resource group on alice. What am I doing wrong? I would
>> assume there must be a way to avoid these.
> 
> does it help to replace the infinity order constraint with a score of 0?

That order constraint is in there for a reason; what should be the
motivation to disable it?

Florian

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 260 bytes
Desc: OpenPGP digital signature
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20090710/e9694945/attachment-0004.sig>