Backing up 20TB Linux VM

[bdoe]

I have several Windows file servers in the 15-20TB range. When I started using Veeam years ago, it was recommended to use spanned disks for these, so Veeam could use parallel processing and back these up faster, which has worked well.

I now have a request for a large Linux server (probably CentOS 8.2) to run an NFS share for client systems. It needs to be a similar size, around 20TB. For the purposes of Veeam (latest 10.x on ESXi 6.5), what should the disk layout be? Should it use a single large vmdk or several smaller ones in LVM?

[nitramd]

I'd suggest several smaller disks to get the benefit of parallelization; of course it depends on your definition of small disk, and ultimately on disk usage.

[bdoe]

I've usually used 5 x 4TB in Windows. I'd probably do similar here, not a ton of disks but enough to let it work parallel.

[Andreas Neufert]

As a single big disk VM is processed slower than VM with multiple smaller disks, you need to find a good balance. I would stay below 10 disks for sure with a sweet spot arround 5-6 disks.

[bdoe]

So backing up LVM works just as well as Windows spanned?

[Andreas Neufert]

From Backup perspective there is no difference (just a VM with multiple VMDKs).

At restore the Linux FLR will support LVM (check documentation about version and supported Linux OS). Setup a small test VM with the LVM configured (small disks). Name everything as you would and see if the Linux FLR works for you and you can restore. (POC test)

[Daniel S]

I have a daily reverse incremental backup job with 70 VMs. Amongst them is a file server - a Linux VM which has a 20TB virtual disk, plus another 6TB virtual data disk. This file server has many GB of changed data per day and my reverse incremental backup job always completes within less than 30 minutes. VM virtual disks are stored on custom-built FreeNAS with 2x vdevs of 11x 7200rpm spinning disks each. Backup storage is simple RAID 6 of 7200rpm spinning disks.

Conclusion: If you want to stick to a single disk for simplicity, backup times should still be reasonable.

[Andreas Neufert]

Correct it is not a matter of backup times as change rate is usually not that high and you will have to backup less than 10% of the data daily.
But it can be painful is you start a normal restore with 100% transport of the data. Instant VM Recovery from Backup or Snapshot help here to address this.