Can't repair cache pool

[bonienl]

This morning I received an email stating the fstrim on my cache failed.

I stopped and restarted the array in maintenance mode to peform a file system check.

It tells there are errors found.

checking extents
incorrect offsets 15700 15647
bad block 7558660112384
ERROR: errors found in extent allocation tree or chunk allocation
checking free space cache
checking fs roots
incorrect offsets 15700 15647
root 5 inode 5639768 errors 400, nbytes wrong
root 5 inode 5671907 errors 2001, no inode item, link count wrong
	unresolved ref dir 257 index 31 namelen 8 name heimdall filetype 2 errors 4, no inode ref
root 5 inode 5688390 errors 2001, no inode item, link count wrong
	unresolved ref dir 7518 index 4935 namelen 46 name f72891df83118db63bb1d2baefe6748e128519.content filetype 1 errors 4, no inode ref
root 5 inode 5691696 errors 2001, no inode item, link count wrong
	unresolved ref dir 7584 index 3125 namelen 46 name ddd6d771f141834a48d34de7040c12cade34d8.content filetype 1 errors 4, no inode ref
root 5 inode 5735288 errors 2001, no inode item, link count wrong
	unresolved ref dir 390 index 40 namelen 21 name 1.12.0.4829-6de959918 filetype 2 errors 4, no inode ref
root 5 inode 5762176 errors 2001, no inode item, link count wrong
	unresolved ref dir 263296 index 17 namelen 32 name 1b021d186d4c5ea6b9bf6522c003fb28 filetype 2 errors 4, no inode ref

Next I run the btrfs --repair option to do a repair, it starts with shifting items but ends with an error (see below)

Shifting item nr 80 by 53 bytes in block 7558660997120
Shifting item nr 81 by 53 bytes in block 7558660997120
Shifting item nr 82 by 53 bytes in block 7558660997120
Shifting item nr 83 by 53 bytes in blocitems overlap, can't fix
check/main.c:3949: fix_item_offset: BUG_ON `ret` triggered, value -5
/sbin/btrfs[0x455e75]
/sbin/btrfs[0x40de1f]
/sbin/btrfs[0x413ca0]
/sbin/btrfs[0x414a62]
/sbin/btrfs[0x415024]
/sbin/btrfs(cmd_check+0x1114)[0x458704]
/sbin/btrfs(main+0x7d)[0x4172ad]
/lib64/libc.so.6(__libc_start_main+0xeb)[0x152cff3e20ab]
/sbin/btrfs(_start+0x2a)[0x4173ba]

I need some expert help to do manual (CLI) repair.

Thanks.

 

[JorgeB]

Basically btrfs --repair should only be used if told so by a btrfs maintainer or somebody else very experienced using it, and I don't consider myself one, if cache is unmountable you can use this to help with backup, reformat and restore, to really try to repair the fs you'd need to ask for help on the btrfs mailing list.

[bonienl]

Cache is still mountabe.

Doing a scrub on cache doesn't return any errors.

I am going to rebuild as you suggest. Thank @johnnie.black

[JorgeB]

Yeah, until there's a functional fsck for btrfs anytime there's fs corruption backup, format, restore should be the default action.

[martinf]

Why is this happening? 1 to 4 times per month, I have this problem. Getting sick of it.

Any way to use xfs rather than btrfs? Have had the same problem with two different ssd, 3 different controllers....

/Martin

[JorgeB]

9 minutes ago, martinf said:

1 to 4 times per month, I have this problem. Getting sick of it.

This usually means there's a hardware problem, like bad RAM, you can try xfs, it's more robust, but if there are RAM errors it will still corrupt data.

[martinf]

2 hours ago, johnnie.black said:

This usually means there's a hardware problem, like bad RAM, you can try xfs, it's more robust, but if there are RAM errors it will still corrupt data.

Think I have a corrypt ram slott. Used 3 sticks. Removed one and are now waiting to see. Will get a new computer when shops are open again.

Cheers