[Linux-cluster] RHEL5 failover domain

[Lon Hohberger]

On Mon, 2007-11-19 at 13:13 +0100, Dariusz Skorupa wrote:
>     I've found problem with RHEL5 cluster. When I use prioritized fail 
> over domain and next reset the node witch have priority set to 1 cluster 
> relocate service to node with priority 2. Next, when node 1 come back 
> ,cluster is trying to relocate service back to primary node.  In logfile 
> I always find:
> 
> Nov 19 12:32:26 l2 openais[1977]: [TOTEM] entering OPERATIONAL state.
> Nov 19 12:32:26 l2 openais[1977]: [CLM  ] got nodejoin message 
> 192.168.10.10
> Nov 19 12:32:26 l2 openais[1977]: [CLM  ] got nodejoin message 
> 192.168.10.11
> Nov 19 12:32:26 l2 openais[1977]: [CPG  ] got joinlist message from node 1
> Nov 19 12:32:26 l2 clurgmgrd[2687]: <notice> Stopping service 
> service:vsftpd
> Nov 19 12:32:41 l2 clurgmgrd[2687]: <err> #52: Failed changing RG status
> Nov 19 12:32:56 l2 clurgmgrd[2687]: <err> #57: Failed changing RG status
> Nov 19 12:32:57 l2 clurgmgrd: [2687]: <info> Executing 
> /etc/init.d/vsftpd status
> 
>    I tested this many times and in this case clurgmgrd do not try to run 
> script with stop parameter, but when I try to relocate service manualy 
> using clusvcadm or when both nodes have priority 1 everything is 
> successful. Is also successful if I'm restarting node using reboot. I 
> think automatic (after crash) relocating starts too early. In my opinion 
> cluster do not wait for rgmanager start.

https://bugzilla.redhat.com/show_bug.cgi?id=327721

-- Lon