Steps required to make use of ReFS?

[Blue407]

Hey All

I have just upgraded my 2012 R2 server to 2016 (in-place upgrade), already running B&R 9.5. Backups are working successfully to local repositories and Tape.

I have 2 existing repositories on this server, using D: and E: drives.

What steps do I need to go through to migrate my existing backups/replications somewhere, re-format the D: and E: drives, and then start being able to leverage ReFS for backups?

[Mgamerz]

I am pretty sure you can just copy the backup files (and the rest of the stuff) to another drive, format as refs, make new repo (it will see it with refs and give refs settings) and then move files back and rescan the repo. That's how I've moved backups between servers, don't think it's different for just moving on-same server.

[nmdange]

Yes the important part is to remove and re-add the repository in Veeam

[Mike Resseler]

The only thing to keep in mind here is that the backup files won't use the block-cloning technology of ReFS before you have written a first full again on that "new" repository and then do synthetic fulls. Just copying data doesn't write the files in the correct format to be ready for blockcloning

[Blue407]

I'm doing Reverse incremental's, does that make a difference?
What should i do in that case?

Could I create a temporary repository somewhere and then use a copy job to re-populate the new ReFS repository?

[foggy]

If you're going to continue backing up locally, backup copy is not an option. Just copy your existing chains back and perform new active full to enable block cloning. Backup method doesn't make any difference.

[Blue407]

Could you please explain the backup copy bit?

At the moment I have full reverse incremental every 2 hrs during the business day, to a dedicated Veeam backup server. Then once a day I use a backup copy job to copy the latest one to an off-site server.

[foggy]

What I meant was the fact that backup copy will not help you in migration of existing backups to ReFS - you just wouldn't be able to map the regular jobs to backups created by it (if that was what you were thinking about).

[dimaslan]

To enable fast clone you could run a new full but you can also run a defrag and compact, that will also enable fast cloning. The thing is, reverse incremental has no meaning when it comes to refs, you will want to do forever incremental with synthetic full (it's done really fast so you can do it whenever, we normally set them up for Sat or Sun). Also, because a synthetic full essentially creates a new backup chain, you can increase the retention without the fear of losing the whole thing due to corruption. Space savings are also amazing, just check the other topics out there that mention the amount of ram you will want to have, this is essential.