[sheepdog-users] run vm on recovering node

Liu Yuan namei.unix at gmail.com
Thu Dec 20 12:24:34 CET 2012


On 12/20/2012 07:15 PM, Liu Yuan wrote:
> Actually the answer is you CAN run the guest from any node when cluster
> is recovering.
> 
> If no object cache enabled, you don't need to do anything special to
> migrate the VM.
> 
> If object cache is enabled, you need to do following before migrating
> the VM:
>  - stop the VM named test
>  - run 'collie vdi cache delete test' which will assure that all the
> data get up to date of particular VM.
> 
> Then you can enjoy running Test any node you want.

Live migration should also work with object cache. But currently, you'll
have to manually remove the cache after VM is migrated.

For example, you want to lively migrate VM named 'test.
 - do live migration using QEMU command on node A
 - after VM is migrated, use 'collie vdi list' to get the VDI ID of
test, for e.g, it might be 'fd32fc'
 - manually remove /store/cache/fd32fc directory. Done.

Thanks,
Yuan




More information about the sheepdog-users mailing list