[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Linux-cluster] problems with gfs and openais cluster.

On 24/08/09 14:41, Steven Dake wrote:
On Sun, 2009-08-23 at 23:28 -0400, Enrique Sanchez wrote:
On Mon, Aug 17, 2009 at 7:06 PM, Steven Dake<sdake redhat com>  wrote:
On Sat, 2009-08-15 at 08:12 -0400, Enrique Sanchez wrote:
On Sat, Aug 15, 2009 at 2:23 AM, Steven Dake<sdake redhat com>  wrote:

The openais.conf file is not used when starting with the cman service.
Instead cman uses its own configuration file
in /etc/cluster/cluster.conf.  openais should not be chkconfig'ed on
unless running without cman.

One reason you could have the error is that you have multiple clusters
on the same multicast address as specified in /etc/cluster/cluster.conf.



thanks for the quick reply: the mcast address is the same and it
looks like it is using the same VLAN here are the details....

this is one cluster (cluster one):

         <cman expected_votes="1" two_node="1">
                 <multicast addr=""/>

and this is in the 2nd cluster (cluster two)......

         <cman expected_votes="1" two_node="1">
                 <multicast addr="" interface="eth1"/>

unfortunately eth1 is on the same VLAN for both clusters...is cluster
1 using the mcast address on both adapters?


clusters are uniquely identified by their multicast address:port pair.
I suggest changing one of the cluster multicast addresses to


I think I need to use the<command>  ccsd -m</command>  to
make the change, just I am not sure whether I need to do a cold start
on the cluster, will this bring the service down or can this be done
concurrently without affecting the service.

That command just changes the multicast address that ccsd uses to communicate. You might need that in addition to the entry in cluster.conf if you have to change the address for ccsd itself. Otherwise just adding the <multicast addr > tag in cluster.conf will work.

If you change the multicast address of a cluster you need to take down all nodes and restart them. Multicast is how they communicate...it's not possible to change without a cluster restart, sorry.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]