[Pacemaker] pacemaker/stonith running "amok"

[Andrew Beekhof]

On Nov 5, 2008, at 6:33 PM, Raoul Bhatia [IPAX] wrote:

> hi,
>
> first off, please find the hb_report at [1].
>
> what i did to my 2 node cluster (wc01, wc02)
>
>> wc02# crm_standby -l reboot -N wc01 -v true
> i verified that wc01 was in standby and (at least i think) the  
> resources
> have been migrated off from wc01.
>
>> wc01# apt-get -u dist-upgrade
> upgraded apache2
>
>> wc01# sync;sync;reboot
> rebootet wc01 as i thought "-l reboot" will make wc01 rejoin after the
> reboot.
>
> wc01 came up but was still considered in standby mode. all of a  
> sudden,
> the cluster continuously rebooted wc02 until i finally moved wc01
> out of standbymode with:
>
>> #wc01: crm_standby -v off -N wc01 -l reboot
>
> can any1 please explain what i did wrong?

The logs don't go back far enough to say.

At 18:18:05 the PE is invoked and sees that wc02 is failed and starts  
to shoot it - but there is no record of it leaving the ccm.
Then all the stonith commands fail - you might want to check the script.

But there is no record at all of wc01 rebooting or wc02's reaction  
when it returns.