[Pacemaker] Trouble getting stonith with external/ipmi to work

[Dejan Muhamedagic]

Hi,

On Fri, Jun 25, 2010 at 11:58:01AM -0500, Bart Willems wrote:
> I am setting SLES11 SP1 HA on 2 nodes and would like to use external/ipmi
> for stonith. I have setup a resource that successfully migrates an IP from
> node1 to node2 when I turn off openais on node1, and migrates back when I
> turn openais back on on node1. Stonith is not rebooting or powering off
> node1 though. The 2 nodes are named storm and storm-b and have the following
> network setup:
> 
> storm:
> ======
> eth0: 172.16.0.1/16 (static)
> eth1: 172.20.168.239 (dhcp)
> ipmi: 172.16.1.1/16 (static)
> 
> storm-b:
> ========
> eth0: 172.16.0.2/16 (static)
> eth1: 172.20.168.114 (dhcp)
> ipmi: 172.16.1.2/16 (static)
> 
> Output of "crm configure show":
> ===============================
> node storm \
>         description="master"
> node storm-b \
>         description="slave"
> primitive publicIP ocf:heartbeat:IPaddr \
>         params ip="143.219.41.20" cidr_netmask="24" \
>         op monitor interval="30s" \
>         meta target-role="started"
> primitive storm-fencing stonith:external/ipmi \
>         meta target-role="started" \
>         operations $id="storm-fencing-operations" \
>         op monitor interval="60" timeout="20" \
>         op start interval="0" timeout="20" \
>         params hostname="storm" ipaddr="172.16.1.1" userid="****"
> passwd="****"
> location storm-fencing-location storm-fencing +inf: storm-b
> property $id="cib-bootstrap-options" \
>         dc-version="1.1.2-2e096a41a5f9e184a1c1537c82c6da1093698eb5" \
>         cluster-infrastructure="openais" \
>         expected-quorum-votes="2" \
>         no-quorum-policy="ignore" \
>         last-lrm-refresh="1277324739" \
>         node-health-strategy="only-green" \
>         stonith-enabled="true" \
>         stonith-action="poweroff"
> op_defaults $id="op_defaults-options" \
>         record-pending="false"
> 		
> I can succesfully reboot the node named storm by issuing the following
> command on storm-b:
> 
> # stonith -t external/ipmi -p "storm 172.16.1.1 **** **** lan" -T reset
> storm
> 
> where the **** are the IPMI username and password. Log file entries on
> storm-b also seem to indicate monitoring is working fine:
> 
> Jun 25 12:53:01 storm-b stonith-ng: [6533]: info: stonith_device_action:
> Operation monitor on storm-fencing active with pid: 10821
> Jun 25 12:53:01 storm-b stonith-ng: [6533]: info: stonith_command: Processed
> st_execute from lrmd: rc=10821
> Jun 25 12:53:01 storm-b external/ipmi[10826]: [10836]: debug: ipmitool
> output: Chassis Power is on
> Jun 25 12:53:02 storm-b stonith: external/ipmi device OK.
> Jun 25 12:53:02 storm-b stonith-ng: [6533]: info: log_operation: Operation
> 'monitor' [10821] for device 'storm-fencing' returned: 0
> 
> Any suggestions on how to get IPMI to turn off storm when the publicIP
> resource migrates would be most appreciated.

Fencing (stonith) is a sort of last resort measure to get the
cluster back to a sane state. Why do you want to turn off a node
just because one IP resource migrated?

Thanks,

Dejan

> Thanks,
> Bart
> 
> 
> _______________________________________________
> 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://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker