[Linux-cluster] clusvcadm -M for VM inside service container

[Dirk H. Schulz]

Okay, I should have tested myself before asking.

It is as suspected below:
A <vm> defined inside a <service> definition can not be migrated live. 
It can be relocated, but that implies stopping and starting.
A <vm> defined "on it's own" can be migrated live (that is using 
clusvcadm -M).

I would have preferred to bundle up several vms to an entity that can be 
started, stopped and live migrated as one. That seems to be impossible, 
then.

Is that correct or do I overlook something?

Dirk

Dirk H. Schulz schrieb:
> Hi folks,
>
> I am new to this list and hope this has not been asked too often, but 
> I did not find many posts on this problem on the web, and none with a 
> solution.
>
> I am running a cluster on CentOS 5.4 with VMs as services, defined 
> like this:
> <service autostart="1" exclusive="0" domain="VirtMach02" 
> name="S-MasseCent5-20" recovery="relocate" hardrecovery="0" 
> max_restarts="5">
>            <vm name="MasseCent5-20" path="/etc/xen/config-san" 
> migrate="live" use_virsh="0" hypervisor="xen" 
> migration_mapping="...,..."/>
> </service>
>
> When I do a live migration using xm it works like a charm.
> When I do a live migration using "clusvcadm -M S-MasseCent5-20 -m 
> /targnode/" I get the following error:
>> Trying to migrate service:S-MasseCent5-20 to targnode...Invalid 
>> operation for resource
> Since clusvcadm -M is restricted to be used with VMs only, I suspect 
> that it does not work for VMs defined inside <service> containers, 
> just for "naked" <vm> definitions.
> Is that correct? Or is there some other reason?
>
> Any hint or help is appreciated.
>
> Dirk
>
> -- 
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster