Another boring question about ZFS on top of HW raid

[jnkraft]

I know it is very controversal subject and generally a bad idea. But maybe i have a little specific case?

Very strict hardware resources and no budget at all.
Old Proliant server, single PVE node, fully filled disk bay, HBA mode, usual raidz2 ready, this is trivial. It has FC HBA card PtP connected to old HPe P2000 Storage. There is no way and no sense to use P2000 any other way than this in my situation - no any other FC-capable hardware in infrastructure.
I want a backup server out of this with bit of non-critical VMs. Idea is to make PBS VM (for PVE Cluster on another hardware), passthrough bay disks to it for normal ZFS "pool 1", and passthrough most of P2000 storage (raid 6) as single disk to make single-vdev ZFS "pool 2" to get dedup/compression/snapshots/etc features that are good for backups storaging.

Please advise me, in this particular case will it be ok to use zfs on top of hardware raid or i'm trying to make something goofy? My guess it would be ok, but there are too much scary warnings in the internet.

 

[Stoiko Ivanov]

Two points up front:
* PBS really profits from fast storage and the performance on older hardware can really be disappointing (starting from AES encryption rates, to slow writes to spinners, to very long verify+garbage collection cycles)
* ZFS behind a RAID controller can cause quite a bit of head-ache and also bad performance (in my personal experience it always has, although I know people who are quite content with their similar setups)

That being said - why not directly install PBS on the old node - and not use ZFS but EXT4/XFS on top of the hardware raid?:
* PBS does the compression (zstd) and deduplication and encryption independent of the used filesystem (it does not use ZFS compression/dedup/encryption)
* PBS does verification to prevent bitrot (also independent of ZFS)

I hope this helps!