Thomas Sjolshagen schrieb:
Keep in mind that the sevice (vm) is considered "up" the moment the
guest starts, ie not booted, so the delay would be minimal.
A more thorough approach to ordering guest start-up and delaying it
(either until some in-guest service is available and/or some time
interval after the previous guest has started would probably require
using your own script and the <script> resource. However, I think that
would mean that you also would need to roll your own solution relative
to live migration, etc since I believe rgmanager only supports live
migration of KVM guests as <vm> resources..?
Well, you're right, vm service must be considered up as soon as "xm
create" returns it's exit code which is very early in the vm startup
process.
Yeah, and your also right with rgmanager supporting live migration only
for standalone <vm> resources, not for <vm> rescources inside service
definitions (and that regardless of KVM or Xen). I have already run into
that.
But what about copying vm.sh to vm-something.sh and filling in a crude
"sleep 60" after the vm start? Would you see any problems there (apart
from updates to vm.sh going past it)?
Thanks for sharing thoughts.
Dirk
// Thomas
On Jan 29, 2010, at 7:47, "Dirk H. Schulz" <dirk.schulz@xxxxxxxxxxxxx>
wrote:
Dirk H. Schulz schrieb:
Hi folks,
the VMs in my cluster are defined via <vm> containers which are not
inside <service> containers because the latter prevents live
migration (as far as I could find out).
Now when I start up the cluster, all VMs are started at once, which
I would like to prevent. Is there any way to define a startup
sequence or startup delay (like you can for the xendomains service)
without misusing dependencies?
Dirk
--
Linux-cluster mailing list
Linux-cluster@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/linux-cluster
Just for the archive: I have an idea now.
I will try defining soft dependencies between the VMs (as I
understand soft is only regarded at startup), so I can have a startup
sequence which is even more than just a startup delay.
Any reasons why this should not work?
Dirk
--
Linux-cluster mailing list
Linux-cluster@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/linux-cluster
--
Linux-cluster mailing list
Linux-cluster@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/linux-cluster
--
Linux-cluster mailing list
Linux-cluster@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/linux-cluster
