Re: Write time tree block corruption detected

[auxsvr@gmail.com]

On Monday, 31 May 2021 00:30:00 EEST auxsvr@gmail.com wrote:
> Hello,
> 
> On linux 5.3.18-lp152.75-default (openSUSE 15.2), I got the following error message:
> 
> [33573.070335] BTRFS critical (device sda2): corrupt leaf: root=3 block=531135201280 slot=0 devid=1 invalid bytes used: have 64503152640 expect [0, 64424509440]
> [33573.070338] BTRFS error (device sda2): block=531135201280 write time tree block corruption detected 
> [33573.071909] BTRFS: error (device sda2) in btrfs_commit_transaction:2264: errno=-5 IO failure (Error while writing out transaction) 
> [33573.071911] BTRFS info (device sda2): forced readonly 
> [33573.071913] BTRFS warning (device sda2): Skipping commit of aborted transaction. 
> [33573.071915] BTRFS: error (device sda2) in cleanup_transaction:1823: errno=-5 IO failure 
> [33573.071917] BTRFS info (device sda2): delayed_refs has NO entry 
> [33577.283856] BTRFS info (device sda2): delayed_refs has NO entry

After using btrfs check --repair /dev/sda2:

ref mismatch on [184344514560 4096] extent item 1024, found 0
owner ref check failed [184344514560 4096]
repair deleting extent record: key 184344514560 168 4096 
Repaired extent references for 184344514560
Dev extent's total-byte(61236838400) is not equal to byte-used(63451430912) in dev[1, 216, 1]

without resolving the issue, as the filesystem would still become read-only, I deleted all snapshots and there hasn't been an issue for a few hours of heavy use. However, btrfs check still returns the error regarding size mismatch and this worries me a bit. Is there an easy way to resolve this or should I recreate the filesystem?

It seems that every time free space is less than 1 GiB and I use zypper, I get a similar corruption message. Thankfully, there hasn't been any data loss so far. There's also a backup of the filesystem right after the first incident, in case it is useful for further investigation.

Regards,
Petros