[Linux-cluster] Re: Nodes show up extremely slowly (or not at all)

[Nikolas Lam]

On Fri, 2007-08-24 at 16:03 +0200, Kadlecsik Jozsi wrote:
> Hi,
> 
> On Fri, 24 Aug 2007, Nikolas Lam wrote:
> 
> > > It turned out to be a switch-related multicast problem: openais used the 
> > > default(?) 239.192.6.53 address, while the switch strangely enough passes 
> > > the 224.0.0.x addresses only. So by setting the multicast address in
> > > cluster.conf explicitly to 224.0.0.3 cured the problem.
> > 
> > I think I might be having similar problems. However I seem to be unable
> > to specify a multicast address manually using conga / luci (the web
> > based tool). 
> [..]
> > Could I please have a look at your cluster.conf?
> 
> Sure, I attached ours. Please note, you must specify the multicast address 
> for cman and every clusternode too and the address must be the same 
> everywhere. (See 
> http://sources.redhat.com/cluster/doc/cluster_schema_rhel5.html.)


Excellent - thanks! Got the cluster's own communications to happen on my
specified addresses, but I can't seem to get the fence_xvmd to use a
specified address. Does anyone have an example cluster.conf for a dom0
cluster where they've specified the multicast IP address for the virt
guest daemon (fence_xvmd) ?

Nik