[Linux-cluster] mount: not in default fence domain

David Teigland teigland at redhat.com
Fri Jul 20 17:00:49 UTC 2007


On Tue, Jul 17, 2007 at 04:36:25PM +0200, Sebastian Reitenbach wrote:
> Hi,
> 
> I am still on my openSUSE 10.2, x86_64, cluster-2.0.0. and openais 0.80.1 
> and kernel version 2.6.20.15.
> 
> I got so far, that my two node cluster tries to start when i issue 
> a /etc/init.d/cman start. When I only start one node, the fenced crashes. 
> But when I start both at the same time, the fenced stays alive, as mentioned 
> in an earlier mail in another thread. After about half an hour, I thought I 
> try mounting the GFS2 partition. The mount process hung on both machines 
> too. Then I killed the cman init script on one machine. The other detected 
> this, but then the mount commands ended too, with the following error 
> message:
> 
> Jul 17 15:28:06 srv5 gfs_controld[22012]: mount: not in default fence domain
> Jul 17 15:28:35 srv5 gfs_controld[22012]: mount: not in default fence domain
> 
> the fenced was running, but no fence_ilo, but I am also not sure, whether 
> this one has to run or not.
> 
> what does the error message above is trying to tell me? Can I fix this, or 
> will it "fix itself", when I am getting the cman script starting correctly?

Try again after you've fixed the nodeids.  If it's a problem, try starting
things up manually (without the init script), and use group_tool to verify
that the nodes have both properly joined the fence domain after
fence_tool join, and prior to the mount. (Follow usage.txt for manually
startup.)

Dave




More information about the Linux-cluster mailing list