I have an unraid server (6.12.8) with 2x 20TB HDDs in an array. One of these is a parity drive, the data drive uses XFS. The primary use for the server is backing up my research and academic data, which occurs via nextcloud, and it also runs a tailscale exit node and pihole. For me, data preservation and reliability is the most important factor. Currently, I have ~10TB of data.

 

I have been reading about the differences between XFS, ZFS, and BTRFS, but am still a bit confused about what might be best for my use case. In brief:

1. Is there any benefit of changing to ZFS in the data drive? This seems to suggest that bitrot protection doesn't happen if using ZFS in the array (but this SpaceinvaderOne video might disagree?), but I don't understand why. If true, it seems like the benefits of ZFS in an array don't apply to my use case.
2. If I do want to change to ZFS or BTRFS in the array, do I just unassign the data drive and then reassign it using ZFS, and it will rebuild from parity? Like the above video, but without the need to reallocate data off the drive first.
3. Am I correct in thinking parity drives insure against drive failure but not data failure, and therefore buying a third drive and using it as second parity drive wouldn't offer any additional data protection. Would this be sensible to do anyway, given I only have 2 drives?
4. Would using ZPool be better than the standard array, as this would provide Bitrot protection? Is converting from the current xfs-with-parity setup to a ZPool doable? Are the difficulties of expanding a ZPool in the future likely to be significant given I'm currently using such large drives?

 

 

FYI, I also have a second off-site backup running ubuntu. This uses ZFS with two drives, has a nextcloud client, and runs periodic cron jobs to backup the copy of the nextcloud directory.

2. Parity can't rebuild a different filesystem.

4 hours ago, turbulent said:

Is there any benefit of changing to ZFS in the data drive? This seems to suggest that bitrot protection doesn't happen if using ZFS in the array (but this SpaceinvaderOne video might disagree?), but I don't understand why. If true, it seems like the benefits of ZFS in an array don't apply to my use case.

In my experience true bitrot is extremely rare, but zfs and btrfs in the array can only detect bitrot, but not repair it, but you would still know which files need to be

replaced.

 

4 hours ago, turbulent said:

If I do want to change to ZFS or BTRFS in the array, do I just unassign the data drive and then reassign it using ZFS, and it will rebuild from parity? Like the above video, but without the need to reallocate data off the drive first.

https://docs.unraid.net/unraid-os/manual/storage-management/#converting-to-a-new-file-system-type

 

4 hours ago, turbulent said:

Am I correct in thinking parity drives insure against drive failure but not data failure, and therefore buying a third drive and using it as second parity drive wouldn't offer any additional data protection. Would this be sensible to do anyway, given I only have 2 drives?

Parity adds redundancy, it's never a backup, dual parity with two drives seems overkill to me, much better to use that drive as an external backup.

 

4 hours ago, turbulent said:

Would using ZPool be better than the standard array, as this would provide Bitrot protection? Is converting from the current xfs-with-parity setup to a ZPool doable? Are the difficulties of expanding a ZPool in the future likely to be significant given I'm currently using such large drives?

It's a valid option, each one has advantages and disadvantages, you can choose according to your priorities, to expand a zfs pool, and until raidz expansion is supported in a year or so, you would need to replace all the disks with larger ones, or add a new vdev with the same number of disks.

 

 

 

 

Thank you! All of this was very informative.

13 hours ago, JorgeB said:

It's a valid option, each one has advantages and disadvantages, you can choose according to your priorities...

 

 

 

 

 

I don't expect my storage requirements to grow that large over the next few years. For my priority (data preservation) - would you recommend ZPool over the current setup?

If you don't require the array flexibility, a zpool would be my recommendation, but keep in mind that unlike the array, you will lose the whole pool if there's a serious filesystem corruption issue, or lose more disks than the redundancy can help with, so don't forget to backup accordingly.

Hi JorgeB,

You mentioned adding a ZFS to an array allows you to discover bit rot.... hate to bug you, but if I add a ZFS or convert all my disks to ZFS, how do I discover it? Does it automatically tell you? do you need to run something to figure it out?

 

From your post, I think I will convert to ZFS if it can at least tell me about the rot. 

 

6 hours ago, Renew said:

but if I add a ZFS or convert all my disks to ZFS, how do I discover it?

By running a scrub, any data corruption would be identified there, and list the affected files.