On Mar 14, 2013, at 4:10 PM, Roman Mamedov [off-list ref] wrote:

- a device (sdg1) was added as a write-mostly member to a degraded two-member RAID1 (md3) and started recovering;
- then it had some I/O errors;

Actually, I'm confused. The UNC errors lines aren't directly tied to sdg, but then later …. (The following lines in [] are one contiguous unedited section from the dmesg, in order but split up.)

[  771.438883] Add. Sense: Unrecovered read error - auto reallocate failed
[  771.438887] sd 10:0:0:0: [sdg] CDB: 
[  771.438890] Read(10): 28 00 00 0c 5a 00 00 00 20 00
[  771.438904] end_request: I/O error, dev sdg, sector 809472

… these are lines clearly indicating it's a Read error with sdg. But why is sdg1, which is rebuilding, reading anything at all? 

So it's actually three things:
1. sdg is reading, why?
2. sdg is encountering UNC read errors, probably bad sectors, why it's new right?
3. md3 might be getting the mirrored copy off sdf, I can't actually tell

[  771.438945] btrfs: bdev /dev/md3 errs: wr 0, rd 9, flush 0, corrupt 0, gen 0

That line is btrfs reporting read errors from md3, not corruptions.

[  771.439011] ata11: EH complete
[  771.439035] btrfs no csum found for inode 31016 start 5668864
[  771.479594] btrfs csum failed ino 31016 off 5668864 csum 2731683453 private 0

What are your btrfs mount options? Are you using nodatasum or nodatacow?

Chris Murphy--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html