3 of my 6 raidz2 drives are marked degraded... I have lost everything ?

[AlcSi]

The strange thing is that I can browse, and even create files from the shell, I can also connect remotely with CIFS (but not NFS ?? not sure there, but my NFS connections to FreeNAS have issues right now... may be linked to the pool issue...).

I assume I've lost everything because it's what I know from RAIDZ2 pools but... how come I see my files ?
Any chance I can actually restore this pool's state ?

What would you do at this point ?
I feel like the slightest error may be the last right now...
(Note that I have backups... but it would save me 10+ hours of work if I could just save this pool instead of restoring from a backup)

Thanks for helping a man in despair :D

 

[Alecmascot]

You need to post all the hardware details and the output from "zpool status" (in code tags for readability).

 

[AlcSi]

Thanks for your answer.

While copying the output of "zpool status" I realized a new line was there about clearing the errors... I did a zpool clear and it all came back to normal.
I had used zpool status numerous times but not after my last reboot.... thanks anyway, it helped :D

 

[Arwen]

Be sure to run a scrub now. It may find other problems.


ZFS is one of the first RAID products that does something different in error handing. For example, a hardware RAID-5 that needs a disk replaced, AND has a bad block on another disk, is a total loss. Meaning full restore from backups.

ZFS with RAID-Z1 where one disk has bad blocks, (meaning it needs to be replaced), but you have other disk(s) with bad blocks, you can still replace the disk with the most bad blocks. You have to use "replace in place", which is using another SAS/SATA port and disk bay for the replacement disk. This allows ZFS to read both the disk to be replaced, (at least it's good blocks), and all the other disks in the case of the bad blocks.

Basically you can have billions of bad disk blocks scattered all over your pool and still not loose a single file. It comes down to WHERE the bad blocks are. As long as a ZFS block stripe has enough data to be good, (even if it lost 2 columns in the case of RAID-Z2), all the data would be recoverable without error.

 

[AlcSi]

Thanks a lot for the details !
I've launched a scrub, it's all good for now, I'll report back if needed.

Final edit, for the record :
The scrub procedure found zero errors, whereas just before I had 3 faulted drives with tons and tons of errors.
Once again, as often I see in this forum, the reason was most probably badly connected SATA cables.
I ordered cables with clips, I hope it will help avoiding this situation in the future.