Veeam with vRDM

[jlester]

We are getting ready to virtualize our e-mail system (Linux with Exim/Courier in Maildir format) as a cluster of VMs. The home directory that contains the mail store will be mounted on at least two VMs as vRDM for use as a clustered/journaled file system. We currently use Veeam setup as a virtual appliance backing up from our fiber-channel SAN to another building's iSCSI NAS box for our other VMs.

How does Veeam handle this type of setup that we plan on using for our e-mail? I know I will have to turn off the snapshot capability since snapshots are not supported on a clustered file system (shared SCSI bus), but will everything else work correctly? Due to the nature of the e-mail storage format we use, a crash consistent backup is fine, only a handful of messages would be affected in the event of a failure. So, will Veeam support vRDM setup with a shared SCSI bus between multiple VMs?

Our current e-mail backup takes hours (days for a full backup ~600GB), so moving this to Veeam with the changed block tracking will save us lots of time if it works as we hope.

Thanks,
Jason

[jlester]

I may have partially answered my own question. I disabled quiescence in a backup job and tried to backup one of the hosts, but still received the error about SCSI sharing. Is there a way to use Veeam to backup a VM without doing a snapshot at all? If not, how are other users backing up clustered VMs using vRDM?

[Vitaliy S.]

Hello Jason,

Yes, snapshot is a requirement for image-level VMware backup and replication, but shapshots are not supported on clustered systems leveraging SCSI bus sharing. As for VMs using vRDMs, Veeam Backup fully supports those, including changed block tracking and all types of restores.

Thank you!

[Gostev]

Vitaliy is correct, unfortunately it is VMware limitation that we cannot do much about (no snapshots support for VMs with disks engaged in SCSI bus sharing).

[jlester]

Thanks, I found that out while reading through some VMware docs. We have done more testing and found that our e-mail server is actually IO bound, so we are going to move it to a new SAN drive that is much faster rather than clustering it.