[DRBD-user] Split brain when starting resources by Pacemaker

Andreas Kurz andreas at hastexo.com
Fri Aug 31 17:25:47 CEST 2012

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


On 08/29/2012 07:00 PM, Carlos Xavier wrote:
> Hi Felix.
> 
> Tank you very much to sharing your point of view, so it seams that the
> temporary solution I took it will be a permanent solution.
> Besides DRBD is being started outside of Pacemaker, Pacemaker is aware
> of DRBD by the device and by the filesystem. I tested stopping DRBD and
> starting it again and the partition got umounted ad mounted again, even
> on reboots :-).
> Although sometime i need to do a resouce cleanup to get all the things
> working as it should.

You _must_ configure fencing in DRBD and STONITH in Pacemaker ... as
discussed quite a lot already ....

Regards,
Andreas

-- 
Need help with DRBD?
http://www.hastexo.com/now

> 
> Kind regards,
> Carlos.
> 
> 
> ----- Original Message ----- From: "Felix Frank" <ff at mpexnet.de>
> To: "Carlos Xavier" <cbastos at connection.com.br>
> Cc: <drbd-user at lists.linbit.com>
> Sent: Wednesday, August 29, 2012 4:34 AM
> Subject: Re: [DRBD-user] Split brain when starting resources by Pacemaker
> 
> 
>> On 08/28/2012 10:55 PM, Carlos Xavier wrote:
>>> Hi.
>>>
>>> Since there was no one hint to solve this trouble, the only temporary
>>> solution I found to allow me to continue with the cluster configuration
>>> was wipe out all the related DRBD configuration from the crm and start
>>> the DRBD resources by the /etc/init.d/drbd init script.
>>>
>>> If someone have another solution I´ll be glad to hear about.
>>>
>>> Kind regards,
>>> Carlos.
>>
>> Hi,
>>
>> I believe the problem is that pacemaker is not yet very well equipped to
>> handle dual-primary setups (at least to my knowledge).
>>
>> In a traditional primary/secondary setup, a node becoming master can
>> fence its peer and give you some safety against split brains. It's much
>> more complex for dual-primary. You would basically need to teach
>> pacemaker to remember which node left the cluster last, and make sure
>> the other node does not get to work standalone.
>>
>> Again, I disbelieve that's currently possible. Maybe there's a simpler
>> solution, though.
>>
>> Regards,
>> Felix
>>
>>
> 
> _______________________________________________
> drbd-user mailing list
> drbd-user at lists.linbit.com
> http://lists.linbit.com/mailman/listinfo/drbd-user



-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 222 bytes
Desc: OpenPGP digital signature
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20120831/fe92d456/attachment.pgp>


More information about the drbd-user mailing list