corrupted pool with 2.1.0-rc5

[mabod]

System information

Type	Version/Name
Distribution Name	EndeavourOS
Distribution Version
Linux Kernel	5.12.2 (zen)
Architecture	x86 (AMD)
ZFS Version	2.1.0-rc5
SPL Version	2.1.0-rc5

Describe the problem you're observing

Today I gave 2.1.0-rc5 a try. I had 2.0.4 installed before. install with dkms was going fine, but after boot one of my pools is throwing an error:

  pool: zstore
 state: ONLINE
status: One or more devices has experienced an error resulting in data
	corruption.  Applications may be affected.
action: Restore the file in question if possible.  Otherwise restore the
	entire pool from backup.
   see: https://openzfs.github.io/openzfs-docs/msg/ZFS-8000-8A
  scan: scrub repaired 0B in 07:35:40 with 0 errors on Thu Apr  1 07:35:41 2021
config:

	NAME                       STATE     READ WRITE CKSUM
	zstore                     ONLINE       0     0     0
	  mirror-0                 ONLINE       0     0     0
	    WD40EFRX-WCC7K6XALFKX  ONLINE       0     0     0
	    WD40EFRX-WCC7K2YAJA69  ONLINE       0     0     0
	  mirror-1                 ONLINE       0     0     0
	    WD40EFRX-WCC7K6YX6SAA  ONLINE       0     0     0
	    WD40EFRX-WCC7K4JKN36U  ONLINE       0     0     0

errors: Permanent errors have been detected in the following files:

        zstore/data/BACKUP/rakete_home:<0x0>

I did not do a feature upgrade yet. Obvioulsy it is an issue with dataset zstore/data/BACKUP/rakete_home. It is not mounted and if I try to mount it I get:

6# zfs mount zstore/data/BACKUP/rakete_home
cannot mount 'zstore/data/BACKUP/rakete_home': Input/output error

How should I approach this now?

[AttilaFueloep]

Most likely this is #11294, so your pool should be fine. I'd just wait for #11300 to land or revert d1d4769.

[mabod]

ok. Or can I just go back to 2.0.4? Will that clear the error?

[mabod]

I downgraded to 2.0.4 and the error is still there. :-(
Does that mean a scrub is needed to fix this?

PS
With zfs 2.0.4 the dataset is mounted it least.

[AttilaFueloep]

Yes scrubbing the pool (maybe twice?) should remove the error, at least it did for me. It won't fix anything though, since there actually is no error to fix, it will just clear the remembered IO error which was there due to a code change.

[mabod]

ok. thanks. May be this question is off topic, but why isnt it possible to do selective / partial scrubs for specific datasets. e.g.
zpool scrub mypool/mydataset
This would help dramatically to fix an issue like this.

[AttilaFueloep]

Dunno, I think in theory it should be possible to scrub only blocks referenced by a certain dataset.

[mabod]

One scrub was enough for me to remove the error message. Thanks. I will close this issue.