ReFS integration with SOBR

[eduzimrs]

Hi,

Guys, i'd like to suggest if possible the ability of SOBR work with fast cloning technology, i talked to the support and i was told that
with a SOBR, it is not currently possible to keep a job restricted to a single extent, thus Fast Cloning will not work as expect.

Thanks.

[Gostev]

I thought this was already implemented, let me double check.

[Gostev]

I confirmed this was implemented in the product a while ago already. When the Data Locality policy is used for the Performance Tier, in case of ReFS resource scheduler will pick the same extent for the new full backup where the previous full backup resides. RTS debug log will reflect this:

Code: Select all

Last full storage repository is preferred because of locality policy and virtual synthetic available.

Can you please update your support engineer as well please? I would normally do this myself, but you didn't share the support case ID

[soncscy]

Perhaps did Support mean that there is always a potential for a placement violation? With a few clients, I saw policy violations happen despite the placement policy because the client had (irresponsibly) decided to ignore low space warnings.

Or @Gostev, do you mean to say that SOBR will always try to match ReFS fulls to the same extent whether or not it has the space for it? Your message says "preferred"; am I being too literal?

[Gostev]

Well, of course when the choice is between not creating a backup at all (due to lack of disk space on the preferred extent) vs. creating backup on another extent, we will always do the latter. Because no one wants to come to perform the restore, only to find out backups for the specific VM were failing in the past days due to lack of disk space on the preferred extent, despite sufficient space available on other. Such behavior would be totally against the very definition of "scale-out"

[Steve-nIP]

It works fine with Data Locality turned on. Of course, if you let the extents run way too low on space, it might place files elsewhere, but that's expected and desired behaviour. Better than failing.

However, there is something surrounding the way Veeam estimates how much space it will need on an extent. I'd suggest opening a case if you want someone to investigate the behaviour in your environment.