VBR 9.5 - REFS

[DaveWatkins]

The easiest way is that you should see [fast clone] in the backup log in the GUI for each disk. You may also want to think about formatting with 64k allocation sizes in future.

You can also check if it's detected the ReFS volume by going into the backup infrastructure tab, selecting properties on the repository and going to the repository tab and hitting the advanced button. If it's detected the ReFS volume the top option (Align backup file data blocks) should be grayed out

[Amarokada]

Hi Dave

As I suspected, it seems these repositories have not picked up on the fact ReFS is available. I'm not seeing [fast clone] anywhere and the align option is not greyed out.

So, what has to be done to get Veeam to recognise the change? I have avoided removing the repository entirely since I have many backups jobs that are configured to use it, and Veeam doesn't let you remove a repository until it is unselected from all jobs.

I have tried putting the repo into maintanence mode and back again, and also attempted to force refresh wherever I can within the repo properties, but it's still not picking up that the volume is ReFS formatted.

Do I literally have to delete and recreate the existing repository (and edit all my jobs across all servers) in order to force it to work?

[Amarokada]

Also, it seems if I am to delete the repository and re-add it, I also have to delete all the existing backup jobs from the database, otherwise it tells me to move all the backups to the alternative repository (which is just a dummy one for now, not able to hold all my backups).

[lorenzo]

I suggest you to have a look to Anton Gostev video: "Ask Gostev #1 - ReFS integration FAQ"

[foggy]

I then formatted the local RAID disk with ReFS (4k allocation size), and then kicked off active-full backups of all my jobs.

Have you created a new repository for this new volume?

[Amarokada]

No I didn't, and am quickly realising that was the silver bullet step that I missed after formatting the drive. However I think I have found an easy way to migrate over to a new repository without having to edit all of my jobs or migrate huge amounts of data around.

1. Create a new repository with name "ReFS Scale-Out Repository Extent 1", and point it to a new folder on the ReFS volume of the same name.

2. Edit my scale out repository that consists of only the 1 previous backup repository extent, add the new ReFS one and remove the old one (without evacuation). Ensure that "perform full backup when required extent is offline" is ticked in advanced settings.

3. Kick off new active full backups of everything. I notice as these complete, the previous restore points on the old extent are removed from the database automatically (now shows only 1 restore point for each VM). All VMs warn that CBT data is out of date, but that is to be expected, and it is reset by Veeam.

Once I have a full weeks worth of backups I will delete the old extent folder.

[veremin]

Looks like a valid approach indeed. Kindly, let us know how well it goes. Thanks.

[bryancc]

I am re-configuring my backup jobs with Veeam 9.5 and a remote repository running Windows Server 2016 / REFS. I'm having trouble finding the best path to take with the "transform previous backup chains" setting. In some environments will spaceless full backup technology provide enough space savings to where it might not be necessary to use the "transform previous backup chains" setting anymore? Thanks!

[foggy]

Yes, since subsequent full backups will not occupy additional disk space, the total disk space taken by the entire backup chain will be comparable to the chain with transformation to rollbacks enabled.

[wasc]

Is there a list anywhere of the backup jobs which benefit from the fast clone technology?

For example I assume a Daily forward incremental backup job which has the 'Create synthetic full backups periodically' checked would benefit from REFS, but what if the same job had 'Create active full backups periodically' instead of 'synthetic full'???

Would they both create a full backup that is spaceless and created quicker?

Am I correct in saying as well that reverse incremental jobs benefit from REFS but only to reduce IO and not to reduce disk space (since all the blocks fast clone is moving out of the VBK are being replaced with new blocks)?

It would be nice if there was a table of different types of jobs, and then ticks to say whether each job benefits from a speed boost, but also whether it benefits from spaceless?

[veremin]

Have you checked already the list of operations benefiting from fast clone technology? Thanks.

[yasuda]

It just feels like it's getting harder to justify having the extra ZFS layer in there.

Tom, ZFS provides bit rot correction; would that make Integrity Streams unnecessary and be a justification for ZFS? I don't know about performance, but ZFS has a longer track record, and maybe there's a benefit to abstracting that from the Windows OS.

[foggy]

For example I assume a Daily forward incremental backup job which has the 'Create synthetic full backups periodically' checked would benefit from REFS, but what if the same job had 'Create active full backups periodically' instead of 'synthetic full'???

Would they both create a full backup that is spaceless and created quicker?

No, Fast Clone is used during synthetic operations.

Am I correct in saying as well that reverse incremental jobs benefit from REFS but only to reduce IO and not to reduce disk space (since all the blocks fast clone is moving out of the VBK are being replaced with new blocks)?

Reverse incremental is already the most efficient in terms of space usage, since the chain consists of a single full + increments. That said, new full backup file created during compact operation does not require extra space in case of reverse incremental as well.

[btmaus]

What's the recommendation for moving existing Veeam restore points to a new ReFS repository?

[foggy]

Glenn, have you reviewed the recommendation given in the thread above?

[btmaus]

Glenn, have you reviewed the recommendation given in the thread above?

Thanks for the reply foggy. So it looks like the below is what we will need to do when setting up a new 2016+ReFS repository?

have some temporary place to copy off all of their backups, reformat, recreate the repo so that it can be recognized as ReFS, copy all the backups back to new repo, rescan and remap existing backups.

[veremin]

Correct.

[foggy]

Don't forget to perform full backup after that to enable new capabilities.

[WRS2200]

Thanks for the post. This was very helpful.

[btmaus]

I have a "best practce" question when adding a new 2016 ReFS Windows Repository to Veeam.

Which of these options should I enable when presented with the Storage Compatibility Settigs:

Align backup file data blocks
Decompress backup data blocks before storing
Use per-VM backup files

Thanks!

[Gostev]

You will find that the first one is selected and disabled when you target ReFS volume, so you can't touch that one anyway

As far as the other two - there is no sense to decompress when going to ReFS, and per-VM is not ReFS specific setting - so use general considerations there (per-VM will give you a better performance at the cost of some extra disk space consumption, and is always a good idea to have enabled beyond a few hundreds of VMs per job for scalability reasons).

[tntteam]

Hi there,

That may already been asked, but I'm not able to find the information

Does refs delivers other advantages than full synthetics and data integrity ?

In my case, I'm using forever incremental, and I was wondering if I would gain any benefit, specially on "backup file merge" and "compact backup files" operations ?

Thanks

[antipolis]

you will benefit from refs on any transform operations, this includes file merge and compaction

[tntteam]

Thanks for info

Do I need to make an active full or anything to activate? I just created my refs repo, and I'm moving data from ntfs right now.
Once done, I'll change the backup specification's repository, and I should be good.

I'll report here any changes in the file merge execution time on the backup tonight.

[antipolis]

you can either do an active full and your backup will be "block cloning aware" right away, or you can just activate weekly synthetics

did you make sure to format your refs volume(s) with 64k cluster size ?

[tntteam]

Yes!
Ok, that's not what I wished to hear, but I'll deal with it

[antipolis]

actually I think if you wait till the next backup compaction it will do the job as well, but I'm not 100% on this, maybe someone else can confirm

[FrancWest]

All,

what about Microsoft's support statement that reFS is not supported on iSCSI luns? Anyone experienced issues with this?

See here:

ReFS is supported with Storage Spaces, Storage Spaces Direct, and non-removable direct attached drives. ReFS is not supported with hardware virtualized storage such as SANs or RAID controllers in non-passthrough mode. USB drives are also not supported.

source: https://docs.microsoft.com/en-us/window ... s-overview

Franc.

[foggy]

It doesn't actually say iSCSI is not supported.

[FrancWest]

Not? How do you read it then?