HA Virtualization with SLES

I have a dream, and the dream is to have an HA cluster running SLES as the virtualization host much like XenServer has for their HA … mainly that if a host dies the VMs on that host get automatically restarted on the other host.

Is this possible? I can’t really find much on this so I’m looking, first, at the “is this possible” side of things before I ask about the “how would you do this” sort of question.

Thanks!

Hi boblmartens,

[QUOTE=boblmartens;17822]I have a dream, and the dream is to have an HA cluster running SLES as the virtualization host much like XenServer has for their HA … mainly that if a host dies the VMs on that host get automatically restarted on the other host.

Is this possible? I can’t really find much on this so I’m looking, first, at the “is this possible” side of things before I ask about the “how would you do this” sort of question.

Thanks![/QUOTE]

you can do this using Pacemaker (from the HAE add-on), where you can have the VMs started on the various hosts of your cluster. We’re running such a setup at the moment.

A major draw-back is the lack of Pacemaker to handle online relocation of dependent resources: If you have two VMs that depend on each other (i.e. a DBMS VM and a corresponding application VM) and you have defined that relationship, the applVM will get stopped when you live-migrate the dbVM to a different server - although you have defined “live migration” as possible for both VMs. So far, we’re living with that and avoid migrating “intermediate” VMs live :wink: Nothing we can do about it, it’s a design flaw of Pacemaker, the developers know about it but seem to be unable to find a proper way to handle this.

Regards,
Jens

Is that sort of thing pretty easy to setup? I guess the best way for me to find out is to try it out on our current VM system.

The shortcoming you’ve identified would be acceptable I feel, at least for us.

Hi boblmartens,

I wouldn’t say so - clusters are easy to start up, but as easy to drop into some hole in the ground that you didn’t notice in the beginning.

Just go for it. Prepare to do some reading to get the fundamentals straight (there’s a good guide for SLES HAE on the SUSE servers) and have a test bed. At least to me, there were a number of options and scenarios I wanted to have a closer look at, before putting things into production.

Once you know about it, you can handle it. It was driving us nuts until we could prove that it wasn’t our configuration’s fault :wink:

For HAE-specific question, i.e. when you set things up, see the SLES HAE forum 'round the corner :slight_smile:

Regards,
Jens

Hi *,

for the sake of those passing by this thread via some search: A few days ago, code was committed to the head of Pacemaker development to overcome the “live migration of dependent resources” issue. It’ll likely be some time until it hits the production packages, but there’s light at the end of the tunnel!

Regards,
Jens

[QUOTE=jmozdzen;18776]Hi *,

for the sake of those passing by this thread via some search: A few days ago, code was committed to the head of Pacemaker development to overcome the “live migration of dependent resources” issue. It’ll likely be some time until it hits the production packages, but there’s light at the end of the tunnel!

Regards,
Jens[/QUOTE]

That’s really cool. Can’t wait to see that make its way down the pipe!