Replace disks and resilver OR Recreate new Pool?

[eingemaischt]

Hi,

I do have a mirrored RAID with 2x2TB drives. Both drives seem to have first problems so I am planning to replace both. I am using Truenas core - but do you upgrades the pool is "legacy encrypted".

So I am wondering if I should add the two new disks to the mirrored raid and remove the old ones after resilvering
or
should I create a new pool with "new" encryption and transfer the fs by zfs send / receive?


Is the 'new' encryption more performant or does it have other advantages?

 

[mistermanko]

should I create a new pool with "new" encryption and transfer the fs by zfs send / receive?

This, as long as the existing pool is in the healthy state. But given that you stated "Both drives seem to have first problems", your pool may be already compromised.
Data on your mirrored pool could already be corrupt or missing, which would make copying over to the new pool a horrendous endeavor.

What does zfs status say?

As to your question about the "new" encryption method, legacy encryption/GELI encryption was the standard up to 11.3. With TrueNAS core 12 zfs encryption became the standard. AFAIK, it is possible to use legacy encrypted pools in TN12 and further, but you may miss out on future zfs flags/features because updating legacy pools is not supported.

 

[eingemaischt]

I shutdown the NAS because I wanted to check RMA status of the hdds - but in the GUI the pool was stated as "healthy".

I heard 'interesting' noises from one of the hdds and checked smartctl - both hdds have more relocated sectors than the threshold.

Does zfs send/receive copy all (sub)datasets and snapshots recursively?

 

[jgreco]

I shutdown the NAS because I wanted to check RMA status of the hdds - but in the GUI the pool was stated as "healthy".

I heard 'interesting' noises from one of the hdds and checked smartctl - both hdds have more relocated sectors than the threshold.

A pool is "healthy" when ZFS hasn't detected any problems with it. This is why it's worth doing both belt and suspenders. SMART doesn't pick up data corruption but ZFS does. ZFS doesn't pick up on reallocated sectors but SMART does.

 

[eingemaischt]

I know. As I stated out: The GUI said that the pool is healthy. Therefore I think that zfs status would report healthy as well, am I right?
I don't want to boot the nas up only to do an unnecessary cli command.

If the "new encryption" would be better an option for me would be to add the new hdds to the old raid but migrate afterwards (through hdds from my work for example). Or to stay with the GELI encryption if it does not affect the performance....

 

[joeschmuck]

It's my opinion that you replace your failing hard drives before doing anything else. Never do two things at once if you can help it. I recommend that you backup you data before doing anything, then you can 1) Destroy vdevs, 2) Replace failing hard drives with new drives, 3) Recreate vdevs, 4) Copy your data back to pool. This action will leave you with the current encryption scheme.

 

[eingemaischt]

I do have regular backups of my "important" data at my neighbor's home ;-) So I am not really afraid. I habe 4 bays and only 2 used - so I can run all hdds for some time in parallel - the new ones and the old ones.

Because the ZFS is a mirrored one I would be able to add 2 new hdds to the existing ones and even keep redundancy through the proces....