Effect of svmotion on jobs based on datastore

[collinp]

What happens if you svmotion a vm to a different datastore and your jobs are backing up at the datastore level? Will the backup job of the destination datastore pick up the new vm? If so, when it is time to do a restore, your vm has now moved to a new job, or potentially many new jobs as the vm migrates to different datastores over time. I assume this would be ok, but it would take some detective work in figuring out which tapes need to be loaded into your tape library to do the restore since the vm could have moved to different jobs over time.

[foggy]

Collin, yes, migrated VMs are automatically picked up and backed up by the corresponding backup jobs. To avoid the work required to keep a log of storage vmotions, you could use jobs based on something else than datastores (folders or resource pools, for example).

[collinp]

Thanks for the info. I am trying to avoid backing up 2 vm's on the same datastore concurrently. I was doing "job chaining" but I guess this isn't the recommended method any more. Chaining also doesn't allow me to do automatic retries on the jobs that are chained. On some of our datastores, there isn't enough free space to perform 2 snapshots at the same time and I also think it would be a lot of I/O. I thought if we backed up by building our jobs around the datastore, we can make sure that 2 jobs aren't running at the same time thus the same datastore wouldn't be hit at the same time. But I think backing up per datastore may prove to be burdensome. The jobs would be much smaller than I would like using this route. Any idea on the best way to accomplish not hitting the same datastore with concurrent jobs on a backup? I have 4 jobs with 20 vm's in each.

Thanks again.

[foggy]

You could use manual datastore selection under proxy settings to limit the particular proxy to access only specific datastores and assigning this proxy to the corresponding job.

[Gostev]

Actually, that's not correct. Manual datastore selection does not work this way, it does not limit proxy to specified datastores - but just points proxy to datastores that our logic may have failed to autodetect.

[foggy]

My bad. But still the different datastores/different proxies approach itself might help in this case.