On Nov 23, 2007, at 1:48 PM, Dejan Muhamedagic wrote:
Hi,
On Thu, Nov 22, 2007 at 05:05:48PM +0530, [EMAIL PROTECTED]
wrote:
looks like I'm barking up the wrong tree. Stonith isn't what I'm
looking for , or is it?
Pls let me explain my scenario...
1. I've a 2 node cluster running HA 2.0.8 .
2. I've 1 resource group with 2 resources : A. a cluster IP
(ocf::Ipaddr) B. a cluster process (ocf::myOcf)
3. Preffered node is active but resource stickiness is +INFINITY
for both. So at heartbeat startup active runs the resource and
standby is in passive state.
4. Whenever, active goes down (network down, m/c crash, heartbeat
stopped) cluster fails resource over to standby which runs it even
when active comes backon (failback=off ; though sometimes failback
does occur )
Split brain problem:
Once, I'd pulled the network cable on active, resource failed over
to standby, then I put the cable back in active, however
accidentally the standby cable got disconnected. After sometime,
with both cables connected, the resource was running on both
machines ( Split-brain I suppose). I now need to fix this, so that
one of the nodes stops running the resource.
Short of having a working stonith device to deal with split-brain
there's nothing which can help you in this situation.
Andrew: Should you perhaps restart the CRM on split brain?
no.
node suicide is a terrible substitute for stonith because it does not
prevent the resources from being started on both machines.
in any event, how would you decide which side to reboot?
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
