Login | Register For Free | Help
Search for: (Advanced)

Mailing List Archive: Linux-HA: Pacemaker

Live migration question.

 

 

Linux-HA pacemaker RSS feed   Index | Next | Previous | View Threaded


lostogre at gmail

Jul 11, 2012, 9:26 PM

Post #1 of 3 (350 views)
Permalink
Live migration question.

I have two cluster nodes that have a gigabit network between them for doing
live migrations of running kvm VMs. If one of the two hosts go off line,
naturally all of the guests then get restarted on the other host. But when
the offline host then comes back online, all of the guests that were
restarted on the "online" host then try to do a live migration. Given that
I only have one gigabit link for doing the transfers, this creates a log
jam. The result is that the VMs that timeout then do a "move" or rather
shutdown VMs to restart them on the newly online node. For my Linux guests,
this is annoying, but with a Windows VM it is a disaster, locking a very
impatient dept out of their server for 3-4 minutes. ( One might think that
this is a miracle, given that before I set up the cluster a server reboot
would take ten minutes. Sigh. )

I would like to make the migrations sequential so that the Windows VM can
migrate first, and then the next most important Linux VMs, etc. Is there
any way to do this?

Beekhof suggested that there were several alternatives and that the mailing
list would be the best place to ask.

lostogre


andrew at beekhof

Jul 29, 2012, 8:02 PM

Post #2 of 3 (300 views)
Permalink
Re: Live migration question. [In reply to]

On Thu, Jul 12, 2012 at 2:26 PM, David Pendell <lostogre [at] gmail> wrote:
> I have two cluster nodes that have a gigabit network between them for doing
> live migrations of running kvm VMs. If one of the two hosts go off line,
> naturally all of the guests then get restarted on the other host. But when
> the offline host then comes back online, all of the guests that were
> restarted on the "online" host then try to do a live migration. Given that I
> only have one gigabit link for doing the transfers, this creates a log jam.
> The result is that the VMs that timeout then do a "move" or rather shutdown
> VMs to restart them on the newly online node. For my Linux guests, this is
> annoying, but with a Windows VM it is a disaster, locking a very impatient
> dept out of their server for 3-4 minutes. ( One might think that this is a
> miracle, given that before I set up the cluster a server reboot would take
> ten minutes. Sigh. )
>
> I would like to make the migrations sequential so that the Windows VM can
> migrate first, and then the next most important Linux VMs, etc. Is there any
> way to do this?
>
> Beekhof suggested that there were several alternatives and that the mailing
> list would be the best place to ask.

You could try creating an ordering constraint between the two VMs, or
setting batch-limit really small.
I think there were some other options too.

_______________________________________________
Pacemaker mailing list: Pacemaker [at] oss
http://oss.clusterlabs.org/mailman/listinfo/pacemaker

Project Home: http://www.clusterlabs.org
Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
Bugs: http://bugs.clusterlabs.org


lostogre at gmail

Jul 31, 2012, 10:43 AM

Post #3 of 3 (299 views)
Permalink
Re: Live migration question. [In reply to]

I'll try those; thanks.

d.p.

On Sun, Jul 29, 2012 at 9:02 PM, Andrew Beekhof <andrew [at] beekhof> wrote:

> On Thu, Jul 12, 2012 at 2:26 PM, David Pendell <lostogre [at] gmail> wrote:
> > I have two cluster nodes that have a gigabit network between them for
> doing
> > live migrations of running kvm VMs. If one of the two hosts go off line,
> > naturally all of the guests then get restarted on the other host. But
> when
> > the offline host then comes back online, all of the guests that were
> > restarted on the "online" host then try to do a live migration. Given
> that I
> > only have one gigabit link for doing the transfers, this creates a log
> jam.
> > The result is that the VMs that timeout then do a "move" or rather
> shutdown
> > VMs to restart them on the newly online node. For my Linux guests, this
> is
> > annoying, but with a Windows VM it is a disaster, locking a very
> impatient
> > dept out of their server for 3-4 minutes. ( One might think that this is
> a
> > miracle, given that before I set up the cluster a server reboot would
> take
> > ten minutes. Sigh. )
> >
> > I would like to make the migrations sequential so that the Windows VM can
> > migrate first, and then the next most important Linux VMs, etc. Is there
> any
> > way to do this?
> >
> > Beekhof suggested that there were several alternatives and that the
> mailing
> > list would be the best place to ask.
>
> You could try creating an ordering constraint between the two VMs, or
> setting batch-limit really small.
> I think there were some other options too.
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker [at] oss
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org
>

Linux-HA pacemaker RSS feed   Index | Next | Previous | View Threaded
 
 


Interested in having your list archived? Contact Gossamer Threads
 
  Web Applications & Managed Hosting Powered by Gossamer Threads Inc.