Local Repository lost in Scale Out Repository with Capacity Tier

[derrick.ritchie]

Good Morning Everyone,

I was going through some hypothetical situations and wondered if this was possible, I have checked the support documentation but do not know if this will work. I hope it will so hoping someone may have gone through it.

We have a scale out repository that contains both a local repository and a object repository. We have a few months worth of full backups in our object repository that have been moved from the local storage. The local storage is pointing to a NAS device on prem. If we loose the local storage entirely, we have an issue where multiple disks fail for example, and we need a new storage device we will need to create a new set of local repositories once this has been configured.

My question is can I use the existing object storage repository in the new scale out repositories if it points to the same backup jobs as before? Will Veeam effectively pick up where it left on in regards to the data in object storage, identify the backups that already exist and progress from there? Or is it best practice to create a new object storage location when you have new local repositories associated with it in a scale-out?

I believe you should be able to use the sync function - https://helpcenter.veeam.com/docs/backu ... ml?ver=100 however I do not want the actual backups downloaded, just the metadata details.

Many Thanks for your assistance.

[Gostev]

Hi, Derrick!

SOBR supports multiple extents in the Performance Tier. There's also the Perform full backup when required extent is offline option that controls the behavior when some of those extents are lost. If you select this checkbox, then losing local extent will not affect backup jobs, as they will simply create full backups to the newly added local extent, and continue on as if nothing happened.

Thanks!

[oleg.feoktistov]

Hi Derrick,

I believe you should be able to use the sync function - https://helpcenter.veeam.com/docs/backu ... ml?ver=100 however I do not want the actual backups downloaded, just the metadata details.

That's actually the case - during synchronization backup files being downloaded consist of metadata only. Data blocks remain in the cloud.

Thanks,
Oleg

[derrick.ritchie]

Appreciate the responses and thank you for the information. When we create the new SOBR and attach the existing Object Repository we will let the sync function download the required data. This should allow the backups to progress once the sync has completed.

[oleg.feoktistov]

If VBR server remains available and only local storage fails, why create another SOBR? Just add new repositories to the existing one as extents. Cheers!

[Gostev]

BTW, I am not 100% sure backup jobs will be able to continue just based on metadata-only stubs. I know all restores will work fine from those, but not so sure about backups. It would be better to re-confirm this with QC, Oleg.

[oleg.feoktistov]

Re-checked with QC and emulated this behaviour in my lab.
If we are removing failed performance extents and adding new ones, here are the points to highlight:

- After re-sync, on the next job run active full is created, thus starting active backup chain.
- If after re-sync you copy inactive backup chain to performance tier, next job run continues exactly as scheduled (I saw increments completed successfully in this scenario).

It's also essential to delete missing restore points from backup primarily to job start.

Cheers!
Oleg

[Gostev]

OK, that's what I thought - we can't continue a backup chain that consists of stubs. In that case, just having jobs do an active full is much faster and cheaper (and cleaner) than downloading backups chain from the capacity tier. Thanks for checking!

[mcz]

Hey guys,

I just read this thread and thought about the restore operation... Suppose you loose your on-prem environment or it isn't available and you'd have to restore. Now as we know veeam can pull the data from the object storage but it only does when it has to, which means if you're having identical blocks on prem, it would read it from there (but not the case in our example). So veeam streams the data from the object storage (for instance in the cloud) and suppose for some reason we were able to bring our on prem environment back to life. Does veeam now realize that and starts reading the blocks from there or is it just "blind" and checks that only during startup of the restore session?

[Gostev]

Hi Michael, we will work with what is available during startup, as this is when all the resources required to perform the restore are determined and tasks are assigned to data movers. Thanks!

[mcz]

Thanks for this very useful statement, Anton. Any plans to check on-prem availability periodically in future releases?

[Gostev]

Honestly, it is very unlikely we will ever go this path, because it requires quite complex logic - so costly to implement, bug prone and with many reliability implications in a real disaster scenario. Plus, all of that complexity is required for a very corner scenario, so it adds minimal value to the product - making it hard to justify.

As a simple workaround, whenever the source infrastructure comes back online, customer can always just stop and restart restore processes from local backups. It is unlikely that anyone will want to continue restoring from the cloud anyway, when the source infrastructure is back online.