[Linux-cluster] tko/interval and emc powerpath multipathing.

Katriel Traum katriel at penguin-it.co.il
Mon Dec 25 15:26:10 UTC 2006


Hello.
I have a 2-node cluster running rhel4 update 4 and cluster suite u4.
The 2 nodes are each connected to an EMC using powerpath for multipathing.
The powerpath failover timeout cannot be changed.
I've set the quorumd tko/interval to several different combos like:
tko=120, interval=1
tko=40, interval=3

Still, when unplugging one of the fiber ports, qdiskd evicts itself
after about 25 seconds:
Dec 19 15:06:52 node1 kernel: qla2400 0000:04:00.0: LOOP DOWN detected (2).
Dec 19 15:07:16 node1 kernel: CMAN: Being told to leave the cluster by
node 2
Dec 19 15:07:16 node1 kernel: CMAN: we are leaving the cluster.
Dec 19 15:07:16 node1 kernel: WARNING: dlm_emergency_shutdown
Dec 19 15:07:16 node1 kernel: WARNING: dlm_emergency_shutdown
Dec 19 15:07:16 node1 kernel: SM: 00000001 sm_stop: SG still joined
Dec 19 15:07:16 node1 kernel: SM: 01000003 sm_stop: SG still joined
Dec 19 15:07:16 node1 kernel: SM: 03000002 sm_stop: SG still joined
Dec 19 15:07:16 node1 clurgmgrd[9357]: <warning> #67: Shutting down
uncleanly
Dec 19 15:07:16 node1 ccsd[7576]: Cluster manager shutdown.  Attemping
to reconnect...
Dec 19 15:07:27 node1 kernel: qla2400 0000:04:00.1: LOOP DOWN detected (4).

Any ideas why this could be happening?

Thanks,
Katriel




More information about the Linux-cluster mailing list