Question regarding direct NFS

[websmith]

Hi,

I am a happy user of Veeam Backup and Replication - and I recently changed my vmware backups to use direct storage access - and my read speeds went from max 100MB/s with nbd to:

Hard disk 1 (50,0 GB) 50,0 GB read at 641 MB/s [CBT] 01:22
Hard disk 1 (32,0 GB) 32,0 GB read at 529 MB/s [CBT] 01:03
etc.

Which is awesome.

But I wonder why the direct NFS has the limitation that no snapshots can be present?

What are the technical reasons why veeam cannot read delta disks?

Is it because it requires the use of the VDDK? or something else?

It would be great if it was possible to backup vm's with snapshots using direct NFS, since nbd is so god damn slow.

Furthermore it would be great to be able to use VMWare Tools quiescence to get fully consistent backups - I have turned quiescence off to gain the speed of the direct nfs access, but it is a shame that if you want fully consistent backups, you need to accept slow backups.

[Gostev]

Hello!

Delta disks use an undocumented format - so while we can read the actual files from the datastore, we cannot make sense of them

VMware Tools provides basic consistency only, without application-specific backup and restore logic. So if you want to get fully consistent backups, you need to enable application-aware processing in the backup job. And unlike VMware Tools quiescence, AAIP is compatible with Direct NFS.

Thanks!

P.S. Please don't keep open snapshots on VMs in any case too many drawbacks and potential issues. Next time, instead of making a snapshot, just trigger a Quick Backup in Veeam! This functionality was specifically added to help customers stop using VM snapshots in production.

[pirx]

As storage snapshots for NFS/NetApp are not supported with Linux proxies I just tested Direct NFS. I also tested this with a VM that has an snapshot.

The job did not complain about the snapshot and finished successfully.

[Gostev]

If you are sure that the backup was performed in [nfs] transport mode, then devs would like to take a look at the logs to see what happened, as this is unexpected and most likely means what was backed up is not an actual VM state, but rather a state as of the time when the existing VM snapshot was created. Please share a support case ID with logs so we could route it straight to the corresponding dev team. Thanks!

[pirx]

Will do so as soon as support portal is working again.

[pirx]

Sorry for the noise, I just realized that the task for the VM with snap silently falls back to nbd mode without any warning.