On Tue, Sep 20, 2011 at 12:40 AM,  <[email protected]> wrote:
> Hi
>
> Ok it was due to the parameter globally-unique which is true by default,
> and that 's lead to the stop of the clone on both sides,

because with globally-unique=true, X:0 on node1 is the same as X:0 on
node2 (but different from X:1 on node2).
so X:0 is active twice which isn't good.  for gfs/dlm/etc,
globally-unique=false is the correct setting.

> whereas when we
> set it to false, there is no more stop on the adjacent healthy node.
>
> But still, it does not work completely, as on the node rebooted, gfs2
> mount
> is stalled until timeout , and then the node is killed again . But it
> seems more
> a gfs2/pcmk problem then a pure Pacemaker problem.

Logs?

>
> Alain Moullé
>
>
>
> De :    [email protected]
> A :     General Linux-HA mailing list <[email protected]>
> Date :  16/09/2011 16:19
> Objet : Re: [Linux-HA] What's wrong in my configuration for GFS2 under
> Pacemaker ?
> Envoyé par :    [email protected]
>
>
>
> Hi ,
> no nothing more with crm_mon -1r
> but I trace in Filesystem script, in fact
> I see that if we configure a clone for fsGS2 (Filesystem)
> it seems that when you ask to start a clone resource,
> Pacemaker at first call the Filesystem stop, and then the
> Filesystem start.
> That's ok if clone is Stopped on both nodes.
> But when you have the clone FS Started on a node,
> and when you start again Pacemaker on the other node,
> when it comes to Start the clone FS, it seems to stop
> it also on the node where the clone FS was always Started,
> disturbing this FS on the node who was always healthy.
>
> Andrew, is this the normal behavior ?
> Or what am I missing on this point ?
>
> Thanks
> Regards
> Alain
>
>
>
> De :    "Ulrich Windl" <[email protected]>
> A :     <[email protected]>
> Date :  16/09/2011 08:04
> Objet : [Linux-HA] Antw: Re: What's wrong in my configuration for GFS2
> under Pacemaker ?
> Envoyé par :    [email protected]
>
>
>
> Hi!
>
> Usually syslog (/var/log/messages, unless redirected) gives you hints of
> what was going on. Also try "crm_mon -1r" when everything is fine, when
> one node is down, when the node is up again. It should help you to learn
> what's going on.
>
> Regards,
> Ulrich
>
>>>> <[email protected]> schrieb am 15.09.2011 um 17:20 in Nachricht
> <ofa545d164.6878f8cd-onc125790c.0053455e-c125790c.00549...@bull.net>:
>> Hi
>> (Pacemaker 1.1.5-5 on RHEL6)
>>
>> As I wrote in my previous msg, I have tested gfs2 under pacemaker,
>> but I 'm facing a problem :
>> When I kill one node (poweroff -f) , the fsGS2 on the healthy node is
>> working fine and IOs are not disturbed.
>> But when the node is rebooted, and I start again Pacemaker,
>> on this node the dlm-clone starts successfully, then the gfs-clone
> starts
>> successfully but when the fsGFS2 is coming : either it fails on the
>> rebooted node, or on the healthy node.
>>
>> (Then if I cleanup the clone-fsGFS2 and start it again, everything works
>
>
>> fine)
>>
>> Has somebody already faced this problem ?
>> Is there a solution in pacemaker configuration itself ?
>>
>> Thanks a lot
>> Regards
>> _______________________________________________
>> Linux-HA mailing list
>> [email protected]
>> http://lists.linux-ha.org/mailman/listinfo/linux-ha
>> See also: http://linux-ha.org/ReportingProblems
>>
>
>
>
>
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
_______________________________________________
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