Hello Michael, thanks for the reply. With one ping_group things now work fine. Maybe somebody could update the wiki page for ping_group and point to pingd instead of ipfail for v2...
I'm still having an issue: as said, with one ping_group things work as expected now (I added rsc colocation rules depending on the attribute "pingd", as described on the pingd wiki page). When using two ping_groups (one for ethernet and one for IB connectivity), I'd expect pingd to failover resources when any of the ping groups fails. But it doesn't (though it reports that one of the groups is dead). It looks like the results of the ping_groups are ORed. Is there a way to change this? Regards, Erich On Montag 02 Juni 2008, Michael Schwartzkopff wrote: > Am Montag, 2. Juni 2008 09:59 schrieb Erich Focht: > > Hello, > > > > I'm trying to get a pair of servers failover when one of their infiniband > > interfaces fails. I hoped to get this functionality with ping_groups and > > added to ha.cf the lines: > > > > respawn hacluster /usr/lib64/heartbeat/ipfail > > ping_group ibping 10.3.0.229 10.3.0.230 10.3.0.231 > > > > Now when I unplug the IB cable on one of the servers I see: > > > > heartbeat[13517]: 2008/06/02_09:39:20 WARN: node ibping: is dead > > heartbeat[13517]: 2008/06/02_09:39:20 info: Link ibping:ibping dead. > > crmd[13533]: 2008/06/02_09:39:20 notice: crmd_ha_status_callback: Status > > update: Node ibping now has status [dead] crmd[13533]: 2008/06/02_09:39:20 > > WARN: get_uuid: Could not calculate UUID for ibping > > > > but nothing actually happens (no failover of resources). > > > > Did I overlook something? > > > > Regards, > > Erich > > HI, > > 1) pingd replaces ipfail completely. No need for ipfail any more. > 2) A ping_group shows "OK" if any of the ping nodes is available. Is this is > behaviour you want? Or do you just want to define 3 ping nodes and failover > if one of the nodes is not available any more? Or even better: Make the > ressources run there, where most of the ping nodes can be seen? > > Config see: > http://www.linux-ha.org/pingd > > Further questions are welcome. > > Back to your logfiles. The cluster perfectly detected that the ping nodes are > not available any more. Did you define any constraints to tell the cluster > what to do, if the connectivily is lost? > > Greetings, > _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
