Hardware RAID and PBS storage
Proxmox Backup Server works best when ZFS can talk to the disks directly. That matters more on a backup box than on a throwaway VM host, because the whole point is being able to trust what is on disk when something has already gone wrong.
With ZFS on PBS, you get checksums, snapshots, and pool management from the Proxmox GUI. That makes day-to-day admin simpler, and it gives ZFS the information it needs to spot bad data and deal with it when redundancy is there. I would rather have that than let a RAID card hide the disk layout from me and hope for the best.
Hardware RAID gets awkward when the controller is part of the recovery story. If the controller fails, you may need the same model, or at least something very close, before you can get at the data again. That is the sort of dependency I do not want on a backup system.
What hardware RAID can get wrong
The main risk is simple: the controller becomes the single point of failure. If it dies, recovery is no longer just about the disks. It becomes a hunt for matching hardware and compatible firmware, which is a fine way to waste a day.
There is also the question of integrity. Hardware RAID works at the block level, so it does not give ZFS the same block-by-block visibility. If a disk has bad data on it, the controller may still try to rebuild the array from whatever it finds. That is not the sort of confidence I want from storage holding backups.
It is also less flexible. Changing RAID level, adding disks, or reworking the array usually means downtime or a messy reconfiguration. For a backup server, that rigidity is a poor trade.
Bit rot is the boring problem that matters
Bit rot is easy to ignore because nothing looks broken until restore time. That is exactly when you do not want to find out a backup set has gone bad.
ZFS checks every block with checksums, and if it has redundant copies it can repair corruption when it finds it. Hardware RAID does not give you that same protection at the filesystem layer. On PBS, that difference is the whole point.
If the backup target is going to sit there for months or years, you want something that can notice corruption before you need the data back. Otherwise you are just storing a problem for later.
Why ZFS keeps winning here
ZFS has become popular in backup setups because it gives you visibility and control without relying on a special controller. It runs on ordinary hardware, and PBS can manage it without much fuss.
Snapshots are useful for backup operations, and compression can save space without much effort. That is practical value, not brochure material. The appeal is that ZFS gives you a clearer view of what is happening on the disks and fewer moving parts in the middle.
I see a lot of people reach for hardware RAID out of habit. On PBS, that habit usually makes recovery harder, not easier.
What I would keep in mind on PBS
- Pick the right layout: ZFS supports mirror and RAIDZ style layouts. Use the one that fits your capacity and redundancy needs.
- Watch disk health: check drive health and scrub results regularly.
- Back up the backups: a PBS box still needs its own recovery plan.
- Test restores: do not assume a backup is good just because it completed.
- Use compression where it makes sense: it can save space without much penalty.
The useful bit is not that ZFS is fashionable. It is that it gives you a better chance of spotting trouble before a restore becomes a disaster.
Drive passthrough on Proxmox
Direct drive access through passthrough is the cleaner option if you want ZFS to see the disks properly. It avoids the RAID controller sitting in the middle and lets ZFS manage the storage itself.
That said, not every RAID controller supports passthrough, and it can make the setup less tidy if you are trying to share disks between guests. I would only use it if the hardware supports it properly and the rest of the layout makes sense.
If the choice is between a controller doing something clever and ZFS getting the raw disks, I know which one I trust more on a backup server.
Basic PBS storage checklist
- Choose hardware that plays nicely with ZFS and passthrough.
- Set up the ZFS pool with the layout you actually need.
- Monitor disk health and pool status.
- Keep a backup schedule in place.
- Test recovery before you need it in anger.
That is about as much glamour as backup storage gets, which is probably a good sign.
