[Linux-HA] placement of stonith resources

Guy Coates gmpc at sanger.ac.uk
Wed Feb 1 04:37:43 MST 2006

> yep :-)
> All resource placement policy is driven from you the admin via the
> constraints and preferences you set in the CIB.

I've tried to avoid the deadlock situations by setting
start_prereq=nothing for the stonith services, so that the stonith
services are always moved to surviving nodes, but it isn't doing what I

I have three nodes, A, B and C.

A is running a shared resource with start_prereq=fencing, and the stonith
resource, with start_prereq=none.

B and C are idle.

If node A fails, I'd expect the following order of events:

1. The CRM to move the stonith service to node B or C.

2. The stonith service then fences node A.

3. On a sucessful fence, the shared resource is started on node B or C.

In reality, the stonith service does not get moved; it is still waiting on
the fencing on node A.

Is this a bug, or have I mis-understood what start_prereq means?



<primitive id="cbi3a-stonith" class="stonith" type="external/ilo-sanger"provider="heartbeat" start_prereq="nothing" on_stopfail="ignore">
           <op id="cbi3a-monitor" name="monitor" interval="120s" timeout="40s" prereq="nothing"/>
           <op id="cbi3a-start" name="start" timeout="40s" prereq="nothing"/>
           <op id="cbi3a-stop" name="stop" timeout="40s" prereq="nothing" on_fail="nothing"/>
             <nvpair name="REALHOST" value="cbi3a" id="cbi3a-stonith-carg1"/>
             <nvpair name="ILOHOST" value="cbi3a-ilo" id="cbi3a-stonith-arg2"/>
             <nvpair name="ILOPASS" value="XXXXXX" id="cbi3a-stonith-arg3"/>

Dr. Guy Coates,  Informatics System Group
The Wellcome Trust Sanger Institute, Hinxton, Cambridge, CB10 1HH, UK
Tel: +44 (0)1223 834244 x 6925
Fax: +44 (0)1223 494919

More information about the Linux-HA mailing list