Best use of vsphere datastore for repository

[ChrisCA]

Hi,

If we've got a vsphere host with a small datastore (2x2tb raid 1) with vms on and a large datastore (4*4tb raid 10) that we want to use for backup storage what's the best way to arrange the repo? At the moment we've had a vm on the small datastore then created virtual disks on the second datastore and added them to the vm and created a veeam repo on each disk. Problem here is -

- inefficient use of storage
- veeam support telling you off for having more than one repo on same vm/host

Should we be having the vm on the first datastore and one large virtual disk on the second datastore, or should we just pass the disk through to the vm?

Thanks,
Chris

[Andreas Neufert]

Hi Chris,
thanks for the questions.

Is it correct that you use the 2x2TB datastore for production VMs? And the 4x4TB exclusive for backups?
Then install the B&R Server on the 4x4TB (100GB boot disk) and a single disk with the rest. Leave some capacity for snapshots on that volume.

In general placing backups on the same server/storage (controller) then your production systems is in general a bad idea.
Please follow: https://www.veeam.com/blog/the-3-2-1-0- ... ility.html

[ChrisCA]

Hi Andreas,

Thanks for the quick reply as ever!

Yes the 2x2tb datastore has production vms on it. As they are only in use during the day and we have a load of jobs that run overnight through our cloud connect server we'd like to put the host to use overnight as a repo (hence the 4*4tb datastore). Don't worry - the production vms on this host are backed up somewhere else!

I want to make sure we're getting the most out of the space on the large datastore, so ideally we'd like a way of using 100% of it as a single repo, and put the repo vm on the small datastore. We're not worried about snapshotting the large disk as we're never going to be backing it up, so is it ok to make a virtual disk that fills the large datastore, or should we pass it through as a physical disk? I guess this is more of a vmware question than a veeam one....

Cheers,
Chris

[Andreas Neufert]

OK, now this is more clear.

It do not really matter were the OS of the Repository Server is running as we do not work with the disks other than the OS and Logs.

I would add the capacity of both in 2 repositories with the "max" capacity of both volumes (beside some space for VMware Snapshots just in case you need those somehow).

Yes there is SOBR to bring both capacity together, but as the 2 extends will have not the same size you will likely fill up the 4x4TB first before you write data to the 2x2TB

[Zew]

Why RAID 10 for backup? RAID 10 should be used for high I/O requirements. RAID 5 would be better suited. just my 2 cents.

[ChrisCA]

Went with raid 10 in the end because we've had issues in the past with raid 5 - long rebuild times, black holes etc.

[Zew]

Sounds like a bad controller IMHO. Yeah rebuild times can be a bit longer, but how often do your discs die? Get disc rot often? (Blackholes?)

It def works but are loosing so much more storage room, which is important on backup repos generally, unless you have a lot of extra disc space then by all means it def will work.

[Delo123]

Always do Raid 10 when somehow possible, period

[foggy]

Why RAID 10 for backup? RAID 10 should be used for high I/O requirements.

And all synthetic backup activity (forever forward, reverse incremental, synthetic backup, all kinds of transform) is extremely I/O intensive.

[Zew]

Good to know, but odd cause normally when people think of long term backup storage they think RAID5, or max storage low I/O....

[Delo123]

If i'm not mistaking he is not talking about archiving here but active backups (also with the potential need to run restores, instant recoveries etc...), I/O will actually be quite high and Harddisks are simply getting too large for raid5, one might just not backup at all in that case

[Andreas Neufert]

As always it depends. If you have the cheapest raid controller in the market with 4 disks ... Raid 10 is maybe a better idea as of raid5/6.

When you have storage systems that spread the load accross many disks (virtualization) Riad 5/6 (or better say 50/60) is more common even as primary backup target.

There are a lot of storage vendor own calculators that provide more insight about performance in different modes and you can compare.

I normally look into the rebuild times of the raid group and if this is fast (special design of the storage vendor) I go with Raid5 or Raid50 and huge 10k SAS disks.