SOBR does not rebalance on active full - #03810460

[mkretzer]

Hello,

because of our previous REFS issues we split our 600 TB REFS volume into 3 200 TB volumes and created a SOBR. We also planned to do periodic active fulls to make sure the SOBR is nicelly balanced (as all things should be...). According to https://helpcenter.veeam.com/docs/backu ... 4#locality this should work and we should never have to worry about REFS loosing block cloning benefits because of partial/non-fast clones until we really run out of space on all extends.

Until we found out that Veeam handles REFS like a deduplicating storage! That means no matter how unbalanced a SOBR is it will always try to write a new active full to the original extend. For us this is fatal as with time our SOBR gets very unbalanced and now we have no way other than loosing block cloning benefits (which is very very bad) to get it balanced again.

Why in the world would Veeam define a REFS volume as deduplicated when there is no deduplication of any kind enabled? Data locality should be enough to keep block cloning with non-active full backups.

I escalated the case and hope someone can come up with a solution for us...

Markus

[HannesK]

Hello Markus,
do you use synthetic and active fulls?

The primary use case for REFS is incremental forever or synthetic fulls. Because that's where block cloning works.

For active full customers, there is no block cloning (except for the rare use case that one uses synthetic and active full), so NTFS would be the way to go.

Best regards,
Hannes

[mkretzer]

Hello Hannes,

we mainly want to use active full as a way to rebalance SOBR and defragment the backup file every 1-2 months.

Without this rebalancing it will balance when you can least live with it: When one extend is already full.

So yes, we need both.

Markus