Is offloaded backup without local metadata useless?

[mcz]

Hi everybody,

I've got a backup where the vbk has been offloaded to s3. Accidentally I once deleted the dehydrated vbk and the vbm. The next restore operation failed and I found out that as I said the vbk/vbm was missing. I thought, no problem, just do a rescan and it should rebuild the local metadata - it wasn't the case and support told me that I would have to download the whole "thing" to get the metadata back. I was wondering what happens if you remove the backup from the configuration and delete the local metadata - as far as I understood this would cause a desaster because support told me that there is no way to rebuild metadata from s3 (allthough I would say that there should be some metadata, AFAIK that's the index).

Of course, it's that same as with every other (normal) vbk, if you delete it it's gone but wouldn't you offload to move your backups from A to B? If you then need metadata on A to be able to restore from B, that just doesn't really make sense to me. Also, it's quite costly to download a whole VBK if it has 1 TB of size just to regain metadata which only has some MB.

Can please someone shed some light on that, maybe there are some information missing that I'm not aware of.

Thanks!

[Gostev]

Hi, Michael.

Simple SOBR rescan should indeed be sufficient in your case, so please ask them to escalate this to R&D for further investigation on what went wrong in your case.

Thanks!

[mcz]

Thanks Anton, that helps a lot. BTW, case number is 03956219

[veremin]

QA team confirmed that indexes should be downloaded from Cloud Tier during SOBR re-scan. As to backup job metadata, it will be restored from cloud tier during next backup job run.

We've noted this as an enhancement for future product versions.

Thanks!

[mcz]

Vladimir,

I'm a little bit confused... When you say 'indexes should be downloaded', that step does not really import the backup, correct? Restoring metadata by running backup job: How could I link a job to an backup that's not imported yet? And what can I do when the vm to backup doesn't exist anymore?

I think I might miss something...

[veremin]

When you say 'indexes should be downloaded', that step does not really import the backup, correct?

Correct.

How could I link a job to an backup that's not imported yet? And what can I do when the vm to backup doesn't exist anymore?

Let's clarify again for the sake of better understanding: what exacts steps you've taken? how current backup chain looks like? and what you want to achieve?

Having this information should allow us to suggest next steps.

Thanks!

[mcz]

Hi Vladimir,

ok, now I've got a much better understanding of my situation...

I had a backup, where the vbk-content got offloaded to S3 and so only the dehydrated vbk + vbm stayed on prem. I made a stupid mistake and deleted the folder with this two files which of course did not delete the backup from the veeam console. The next restore attempt failed and showed me the message about the missing folder (with vbk+vbm). Rescan of the SOBR (including S3) didn't bring the metadata back so I asked support for assistance.

Now in the meanwhile I also deleted the backup in the veeam console, which meant that I completely wiped it out. Then I did another rescan of the SOBR and the following happened:



So luckily it downloaded the metadata but only because the backup wasn't imported. It looks like that if the backup is within the db and metadata on prem got deleted, it won't download them from S3. Maybe a case for future improvements.

[veremin]

Sure, let's consider it a potential enhancement for future versions. However, I still want to point out that if you manually remove .vbm from performance tier, backup server should try to download it from capacity tier during next job run. Thanks!

[mcz]

Vladimir, which "next job" do you mean? backup? offload?

[veremin]

Backup job. Thanks!