In order to see any benefits from Veeams integration with the Windows Server 2016 ReFS filesystem all fulls and incrementals will have to have been created on the new filesystem. This means that just moving data over won’t mean you immediately see the benefits of fast cloning and spaceless fulls. There are a few extra steps that need to be taken first:
View this post to see some of the benefits that come along with using Veeam and ReFS.
Update: Make sure to use 64K Block Size when formatting the Veeam repository volumes to avoid issues with 4K Block Size and ReFS. Read this post for more information.
Migrating Existing Backups to ReFS Volumes
The first thing to note here is that only new full and incremental backups created on ReFS will benefit from the fast cloning and spaceless full technology that comes with ReFS on Windows Server 2016. This means that once you move your data over to the new ReFS volume you won’t begin seeing any performance or storage benefits until either an active or synthetic full is created and then all old incremental restore points have been merged or deleted and new increments are all created on the new volume.
This can prove to be troublesome if you are coming from a deduplicating storage where all previous fulls and incrementals were deduplicated. By moving them to the new volume all of the data would become rehydrated and most likely blow up your storage.
As you may not have enough storage for the full rehydrated size of all of your fulls, archived fulls, and incrementals, it would most likely benefit you to keep those archived backups on the deduplicating storage or write them off to tape and only migrate the newest full .vbk, incrementals .vib, and .vbk to the new ReFS volume.
Keep in mind that you won’t see the benefits of ReFS until the most recent full and incrementals were created on the new ReFS volume. This means that you will need to have storage for at least 2 full backups and all incrementals. We recommend scheduling the GFS retention to create a full as soon as possible so that the oldest archived full on the new volume can be deleted and free up as much space as possible.
Once the Synthetic full backup and the new incrementals have all been created on the new storage you can delete the oldest archive point from the ReFS volume and all new backups will see the benefits from the net ReFS filesystem.
Seeding Offsite Backup Copy to ReFS Volumes
Seeding a backup to ReFS still has the benefit of decreasing the initial WAN utilization by preventing the need to ever to a full backup over the internet. However, with ReFS even after seeding the backup all backups will have to have been created on the volume they were imported to. This means that you won’t see the
We have been successful using the following process when seeding backups to ReFS(Keep in mind that you will temporarily need storage for 2 full backups and 2 incremental restore points.):
- First perform an initial full backup copy (seed backup) to an external drive at the primary location.
- Once the backup is complete ship the drive off to the secondary location and import the .vbm and .vbk file into the target repository.
- Rescan the new repository from the primary site and those new backups should be added to the configuration.
- From the job used to create the seed edit the target repository and point it to the new repository at the secondary site.
- Use ‘Map Backup’ to choose the newly imported backup from the repository.
- Using the following method we have been successful forcing the backup copy job to perform a full GFS synthetic full:
- In the job settings, change the number of restore points to 2(This is the lowest it will allow you to create.)
- Enable GFS retention and change the schedule for weekly backups to occur during the backup following the next backup.
- For example, if your backups happen nightly and today is Wednesday you would allow the incremental backup to run Wednesday night, then schedule the weekly backup to happen during the backup scheduled for Thursday night. The backup needs to run Wednesday first in order to create an incremental backup so that the job hits its limit of 2 restore points. Then on Thursday, because the job has hit its retention and the synthetic is scheduled it should make the GFS restore point.
- Once the GFS synthetic full has been created you can delete the archived full ‘…_W.vbk’ to free up storage. (You can keep it but this archived full will not benefit from any ReFS spaceless fulls and will utilize your storage until it is deleted from retention.)
Now you can change the retention of the backup job to whatever you would like and any new backups will benefit from the new ReFS filesystem.