On Wed, 2007-11-21 at 13:35 +0100, jr wrote:
> Hi everybody,
> I am currently working on an installation of a failover RHCS with Xen
> guests as Services. Everything works great so far, Xen domains are
> failed over to still running nodes if i switch a node off. Even high
> available bonding integrates nicely.
> The only thing yet is the (live) migration of Xen domain Services. i
> changed the /usr/share/cluster/vm.sh so it reads xm migrate -l, but if I
> do:
You can't migrate it if it's got dependencies in rgmanager, because
other resources don't know how to migrate themselves. For example,
migrating a file system is a very strange prospect; mount/umount won't
work...
* Put your GFS volume(s) in /etc/fstab
* Put the VM at the top level; you shouldn't need to reference
clusterfs.sh in cluster.conf in this case.
So - your configuration should be far simpler when using GFS:
> <service name="VirtServ01" domain="failoverdomain1"
> autostart="1">
> <clusterfs ref="XenDomains">
> <vm ref="VirtServ01"/>
> </clusterfs>
> </service>
...becomes:
<vm name="VirtServ01" domain="failoverdomain1" autostart="1"/>
Obviously, you *cannot* migrate when using non-cluster file systems as
Xen domain backing stores.
-- Lon
--
Linux-cluster mailing list
Linux-cluster@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/linux-cluster
