[6.11.3] Cache disks becomes Read-only

Diags are from v6.11.1 but btrfs is detecting data corruption in one of the cache devices:

 

Nov 11 11:20:45 Tower kernel: BTRFS info (device nvme2n1p1): bdev /dev/nvme2n1p1 errs: wr 0, rd 0, flush 0, corrupt 9, gen 0

 

Start by running a scrub.

Thanks Jorge,

 

As mentioned I had to roll back, I did scrub on 6.11.3 and rebooted after, made no difference. Only solution was roll back.

 

Difficult to say more without the v6.11.3 diags other than pool is also showing issues with v6.11.1, other than the mentioned corruption there's also this:

 

Nov 11 11:20:46 Tower kernel: BTRFS error (device nvme2n1p1): incorrect extent count for 6093506871296; counted 8236, expected 8224

 

Looks like it's just a log tree problem, so it might be fixable by zeroing it, but make sure pool is backed up before trying, with the array stopped:

 

btrfs rescue zero-log /dev/nvme2n1p1

 

Then start the array, if the pool mounts run a scrub and when it ends post new diags.

Thanks Jorge, just to close this one out. Decided to reformat cache drive and reassign to pool. Updated to 6.11.3 afterwards, been rock solid since (5+ days).

Looks like somehow 6.11.3 and 6.11.1 handled this corruption differently (guessing 6.11.3 is less tolerant and only mounted as read-only). 

Cheers,

Luis

Changed Status to Closed

Changed Priority to Other