[Pacemaker] STONITH without mandatory success possible?

[Doug Clow]

I'll do some experiments to see if I can get Corosync more reliable.  I'm using Corosync v1 as part of cman-corosync-pacemaker.  RRP with one port on a switch and the other port on a crossover cable between the two hosts (although technically each port is still part of a vSwitch since its a VMware VM). 

I do have fence-agents installed, but on Centos its missing the the null and ssh devices.  Andrew, if you mean I can have stonith devices without enabling stonith globally thereby making success optional I will try that out.  That would be just what I need.

Thanks all,
Doug


On Jun 20, 2013, at 1:38 AM, Sven Arnold <sven.arnold at localite.de> wrote:

>>> 
>>> echo "0" >  /sys/class/net/virbr0/bridge/multicast_snooping
>> 
>> That results in multicast packets are broadcasted to all bridge ports. I
>> prefer to have igmp querier turned on on a central switch.
>> 
>> 
> 
> I had the impression that the OP used virtual machines on a local (virtual and private) network. I had such a setup with my first experiments and experienced (timing) issues with corosync messages until I turned snooping off.
> 
> But you are right, on my production system I have turned on both multicast_snooping and multicast_querier.
> 
> For reference, there was this bugreport that was helpful for me:
> https://bugzilla.redhat.com/show_bug.cgi?id=880035
> 
> Regards,
> 
> Sven
> 
> 
> 
> 
> _______________________________________________
> 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