[Pacemaker] Dual primary DRBD+OCFS2+XEN+Pacemaker failover issues

Digimer lists at alteeve.ca
Fri Jun 6 04:16:49 UTC 2014


Mostly yes. Many, many people use just stonith and it's enough for 
recovery in most failure cases.

However, personally, I always use (and recommend) PDU fencing as a 
backup fence method. The reason is that IPMI shares it's power with the 
same power that feeds the node itself. So if the node totally loses 
power[1], most systems will lose their IPMI as well. This will mean that 
the surviving node will fail to fence the peer. The survivor can not 
safely assume the state of the peer, so the only safe option is to block.

By having backup PDU fencing, pacemaker can use STONITH levels[2] to say 
"try fencing using IPMI, but if that fails, go and try fencing by PDUs".

Generally, you will want two PDUs and nodes with redundant power 
supplies. If you have a single PDU, and that PDU fails (or the 
UPS/circuit behind it fails), you're going to lose both fence methods. 
By using dual PDUs (and dual UPSes/mains circuits), the failure of one 
power rail will leave IPMI working.

This may seem complicated, but it's pretty straight forward and it is 
what we've used for years successfully (albeit on RHCS, though I've 
tested it successfully with pacemaker 1.1.10+).

digimer

1. https://alteeve.ca/w/AN!Cluster_Tutorial_2#Cutting_the_Power_to_an-c05n01
2. http://clusterlabs.org/wiki/STONITH_Levels


On 05/06/14 11:42 PM, kamal kishi wrote:
> Thank U digimer, will try your solution.
>
> But is IPMI Stonith enough for my configuration??
>
>
> On Thu, Jun 5, 2014 at 10:56 PM, Digimer <lists at alteeve.ca
> <mailto:lists at alteeve.ca>> wrote:
>
>     On 05/06/14 08:09 AM, kamal kishi wrote:
>
>         property no-quorum-policy=ignore
>         property stonith-enabled=false
>
>
>     This is not good. You *must* have tested, working stonith in
>     pacemaker. You already set the fence handler and fencing policy in
>     drbd, but it only works if pacemaker's stonith is configured and
>     working.
>
>     Correct this and see if your problem remains.
>
>     --
>     Digimer
>     Papers and Projects: https://alteeve.ca/w/
>     What if the cure for cancer is trapped in the mind of a person
>     without access to education?
>
>     _________________________________________________
>     Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
>     <mailto:Pacemaker at oss.clusterlabs.org>
>     http://oss.clusterlabs.org/__mailman/listinfo/pacemaker
>     <http://oss.clusterlabs.org/mailman/listinfo/pacemaker>
>
>     Project Home: http://www.clusterlabs.org
>     Getting started:
>     http://www.clusterlabs.org/__doc/Cluster_from_Scratch.pdf
>     <http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf>
>     Bugs: http://bugs.clusterlabs.org
>
>
>
>
> --
> Regards,
> Kamal Kishore B V
>
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org
>


-- 
Digimer
Papers and Projects: https://alteeve.ca/w/
What if the cure for cancer is trapped in the mind of a person without 
access to education?




More information about the Pacemaker mailing list