Best Practices for Changed Block Tracking?

[Kaelex]

Hello guys,

Are there any white papers or documentation available on the best setup of VMs for using Changed Block Tracking?

I'm thinking things like putting the swap on a separate drive that's not part of the backup or something like that.

If there is I'd much appreciate being pointed in their direction.


Thanks!

[Bunce]

I'm currently recreating our backup jobs and thought it worthwhile implementing any additional optimisation practices before starting again.

If we come up with any unique ones, perhaps we could jot them down for others to use.

For example,

• Is there any performance or backup size improvements to gain from relocating the Windows page file to a small dedicated VMDK disk and exclude from backups? (will be easier to do in V5 as I believe per-vm exclusions are available) - ?
• Does anyone bother defragmenting anymore (either in Guest or at vSphere level) and how often, given its likely to affect CBT?
• etc

Cheers,
A

[Vitaliy S.]

Hello,

As far as best setup is concerned, I can recommend to perform these two things in mind:

1. Running defrag and then sdelete inside VM, prior to performing a full backup.
2. Placing a swap file to an separate virtual disk and excluding it from backup jobs.

These steps should allow you to keep the backup files to a minimum size possible with the less time taken for an incremental run.

Thanks!

[Gostev]

I agree that you would get some improvements by doing this, but this is something reasonable to do only in small environments with dozens of VMs. In bigger environments with hundreds of VMs, such micro-management is hardly possible and is typically never done - people just backup VMs the way they are.

[Oletho]

1. Running defrag and then sdelete inside VM, prior to performing a full backup.

Wouldn't that just trigger more changed blocks for CBT?

Ole Thomsen

[Gostev]

This would not matter, because you are going to perform this before full backup.
By the way, this recommendation is good for thick disks only (should not be done on thin-provisioned ones).

[Oletho]

Ah, of course - didn't notice the "full"

Ole Thomsen

[Bunce]

This would not matter, because you are going to perform this before full backup.
By the way, this recommendation is good for thick disks only (should not be done on thin-provisioned ones).

It is actually useful in thin disks when you want to reclaim unused space

If your thin disk has previously been filled up, the disk allocation is fully used. If you then delete much of the data from that disk, the allocation won't change until you:
- Run sdelete
- StorageVMotion (or shutdown and migrate) to a different Lun.

http://www.virtualizationteam.com/virtu ... space.html

(Will) come in useful when doing thin-thin replicas in V5 (before running the first replica ).