Re: Nasty corruption on large array, ideas welcome

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Jan 22, 2019 at 6:43 PM Chris Murphy <lists@xxxxxxxxxxxxxxxxx> wrote:
>
> On Tue, Jan 22, 2019 at 9:41 AM Thiago Ramon <thiagoramon@xxxxxxxxx> wrote:
> >
> > Back again with pretty much the same problem, but now without a
> > reasonable cause:
> > I've bought a couple new 8TB disks, recovered everything I needed from
> > my previously damaged FS to a new BTRFS on those 2 drives (single copy
> > mode), double-checked if everything was fine, then wipefs'd the old
> > disks and added the ones that didn't have any issues previously to the
> > new array and rebalanced to RAID6.
> > Everything was running fine through the weekend and I was about 50%
> > done when today:
> > [  +7.733525] BTRFS info (device bcache0): relocating block group
> > 8358036766720 flags data
> > [Jan22 09:20] BTRFS warning (device bcache0): bcache0 checksum verify
> > failed on 31288448499712 wanted A3746F78 found 44D6AEB0 level 1
> > [  +0.460086] BTRFS info (device bcache0): read error corrected: ino 0
> > off 31288448499712 (dev /dev/bcache4 sector 7401171296)
> > [  +0.000199] BTRFS info (device bcache0): read error corrected: ino 0
> > off 31288448503808 (dev /dev/bcache4 sector 7401171304)
> > [  +0.000181] BTRFS info (device bcache0): read error corrected: ino 0
> > off 31288448507904 (dev /dev/bcache4 sector 7401171312)
> > [  +0.000158] BTRFS info (device bcache0): read error corrected: ino 0
> > off 31288448512000 (dev /dev/bcache4 sector 7401171320)
>
> This is corruption being detected and corrected on those listed
> sectors. As this is a bcache device, it's a virtual sector so it's
> hard to tell if it's coming from bcache itself, or the cache device,
> or the backing device.
>
I was using bcache in writeback mode with my old FS, but I've learned
THAT lesson the hard way. This one was just using writearound, unless
bcache REALLY screwed it up I find it hard that it's the source of the
corruption. There were no read or write errors from bcache since the
time the new array went up, and each bcache* device is just a thin
layer over a whole raw disk now.

>
> > [Jan22 09:21] BTRFS info (device bcache0): found 2050 extents
> > [  +8.055456] BTRFS info (device bcache0): found 2050 extents
> > [Jan22 09:22] BTRFS info (device bcache0): found 2050 extents
> > [  +0.846627] BTRFS info (device bcache0): relocating block group
> > 8356963024896 flags data
> > [Jan22 09:23] BTRFS info (device bcache0): found 2052 extents
> > [  +6.983072] BTRFS info (device bcache0): found 2052 extents
> > [  +0.844419] BTRFS info (device bcache0): relocating block group
> > 8355889283072 flags data
> > [ +33.906101] BTRFS info (device bcache0): found 2058 extents
> > [  +4.664570] BTRFS info (device bcache0): found 2058 extents
> > [Jan22 09:24] BTRFS info (device bcache0): relocating block group
> > 8354815541248 flags data
> > [Jan22 09:25] BTRFS info (device bcache0): found 2057 extents
> > [ +17.650586] BTRFS error (device bcache0): parent transid verify
> > failed on 31288448466944 wanted 135681 found 135575
>
>
> Over 100 generations have passed, and yet it's only finding stale data
> on the desired btrfs byte nr (in btrfs linear space) so it might be
> extent tree corruption again.
>
> It's not possible from the available information to do anything but
> speculate how that much data is being lost or somehow being
> overwritten.
>
>
> > [  +0.088917] BTRFS error (device bcache0): parent transid verify
> > failed on 31288448466944 wanted 135681 found 135575
> > [  +0.001381] BTRFS error (device bcache0): parent transid verify
> > failed on 31288448466944 wanted 135681 found 135575
> > [  +0.003555] BTRFS error (device bcache0): parent transid verify
> > failed on 31288448466944 wanted 135681 found 135575
> > [  +0.005478] BTRFS error (device bcache0): parent transid verify
> > failed on 31288448466944 wanted 135681 found 135575
> > [  +0.003953] BTRFS error (device bcache0): parent transid verify
> > failed on 31288448466944 wanted 135681 found 135575
> > [  +0.000917] BTRFS: error (device bcache0) in
> > btrfs_run_delayed_refs:3013: errno=-5 IO failure
> > [  +0.000017] BTRFS: error (device bcache0) in
> > btrfs_drop_snapshot:9463: errno=-5 IO failure
>
> And -5 I/O error is not a Btrfs error either, it's the detection of an
> IO error from the underlying block device, whether real or virtual.
>
Couldn't figure the source of the -5 either, no kernel logs from
anything byt BTRFS complaining about it. After I umounted the array,
it didn't shown up anymore, and I was able to remount the array with
the skip_bg patch.

>
>
> > [  +0.000895] BTRFS info (device bcache0): forced readonly
> > [  +0.000902] BTRFS: error (device bcache0) in merge_reloc_roots:2429:
> > errno=-5 IO failure
> > [  +0.000387] BTRFS info (device bcache0): balance: ended with status: -30
> >
> > Couldn't check anything even in RO mode scrub or btrfs check, when I
> > unmounted the array I got a few kernel stack traces:
> > [Jan22 13:58] WARNING: CPU: 3 PID: 9711 at fs/btrfs/extent-tree.c:5986
> > btrfs_free_block_groups+0x395/0x3b0 [btrfs]
> > [  +0.000032] CPU: 3 PID: 9711 Comm: umount Not tainted
> > 4.20.0-042000-generic #201812232030
> > [  +0.000001] Hardware name: Gigabyte Technology Co., Ltd. To be
> > filled by O.E.M./H61M-DS2H, BIOS F6 12/14/2012
> > [  +0.000014] RIP: 0010:btrfs_free_block_groups+0x395/0x3b0 [btrfs]
> > [  +0.000002] Code: 01 00 00 00 0f 84 a0 fe ff ff 0f 0b 48 83 bb d0 01
> > 00 00 00 0f 84 9e fe ff ff 0f 0b 48 83 bb 08 0$
> >  00 00 00 0f 84 9c fe ff ff <0f> 0b 48 83 bb 00 02 00 00 00 0f 84 9a
> > fe ff ff 0f 0b e9 93 fe ff
> > [  +0.000001] RSP: 0018:ffffa3c1c2997d88 EFLAGS: 00010206
> > [  +0.000001] RAX: 0000000020000000 RBX: ffff924aae380000 RCX:
> > 0000000000000000
> > [  +0.000001] RDX: ffffffffe0000000 RSI: ffff924b85970600 RDI:
> > ffff924b85970600
> > [  +0.000001] RBP: ffffa3c1c2997db8 R08: 0000000020000000 R09:
> > ffff924b859706a8
> > [  +0.000000] R10: 0000000000000002 R11: ffff924b973a1c04 R12:
> > ffff924aae380080
> > [  +0.000001] R13: ffff924b8dfe8400 R14: ffff924aae380090 R15:
> > 0000000000000000
> > [  +0.000002] FS:  00007f1bd1076080(0000) GS:ffff924b97380000(0000)
> > knlGS:0000000000000000
> > [  +0.000001] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > [  +0.000000] CR2: 0000562d2eb13c10 CR3: 0000000156910006 CR4:
> > 00000000001606e0
> > [  +0.000001] Call Trace:
> > [  +0.000018]  close_ctree+0x143/0x2e0 [btrfs]
> > [  +0.000012]  btrfs_put_super+0x15/0x20 [btrfs]
> > [  +0.000004]  generic_shutdown_super+0x72/0x110
> > [  +0.000001]  kill_anon_super+0x18/0x30
> > [  +0.000012]  btrfs_kill_super+0x16/0xa0 [btrfs]
> > [  +0.000002]  deactivate_locked_super+0x3a/0x80
> > [  +0.000001]  deactivate_super+0x51/0x60
> > [  +0.000003]  cleanup_mnt+0x3f/0x80
> > [  +0.000001]  __cleanup_mnt+0x12/0x20
> > [  +0.000002]  task_work_run+0x9d/0xc0
> > [  +0.000002]  exit_to_usermode_loop+0xf2/0x100
> > [  +0.000002]  do_syscall_64+0xda/0x110
> > [  +0.000003]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
> > [  +0.000001] RIP: 0033:0x7f1bd14bae27
> > [  +0.000001] Code: 90 0c 00 f7 d8 64 89 01 48 83 c8 ff c3 66 0f 1f 44
> > 00 00 31 f6 e9 09 00 00 00 66 0f 1f 84 00 00 00
> >  00 00 b8 a6 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 39
> > 90 0c 00 f7 d8 64 89 01 48
> > [  +0.000001] RSP: 002b:00007ffdb15a75a8 EFLAGS: 00000246 ORIG_RAX:
> > 00000000000000a6
> > [  +0.000002] RAX: 0000000000000000 RBX: 000055df329eda40 RCX:
> > 00007f1bd14bae27
> > [  +0.000000] RDX: 0000000000000001 RSI: 0000000000000000 RDI:
> > 000055df329edc20
> > [  +0.000001] RBP: 0000000000000000 R08: 000055df329eea70 R09:
> > 00000000ffffffff
> > [  +0.000001] R10: 000000000000000b R11: 0000000000000246 R12:
> > 000055df329edc20
> > [  +0.000001] R13: 00007f1bd15e18c4 R14: 0000000000000000 R15:
> > 00007ffdb15a7818
> >
> > Now I'm back in a very similar situation as before, btrfs check gets me:
> > Opening filesystem to check...
> > checksum verify failed on 24707469082624 found 451E87BF wanted
> > A1FD3A09
> > checksum verify failed on 24707469082624 found 2C2AEBE0 wanted
> > D6652D6A
> > checksum verify failed on 24707469082624 found 2C2AEBE0 wanted
> > D6652D6A
> > bad tree block 24707469082624, bytenr mismatch, want=24707469082624,
> > have=231524568072192
> > Couldn't read tree root
> > ERROR: cannot open file system
> >
> > I could do it all again, but first, what can be wrong here? This array
> > was working for some 4 years until it went bad a few weeks ago, and
> > now the FS got badly corrupted again without any warnings. Any
> > suggestions? Bad RAM, SAS controller going bad, some weirdly behaving
> > disk? I need to figure out what can be failing before I try another
> > recovery.
>
> I think it's specifically storage stack related. I think you'd have
> more varied and weird problems if it were memory corruption, but
> that's speculation on my part.

I've done a quick memory test with stressapptest and it was fine, so
if it's the memory it's something very localized.
>
> I'd honestly simplify the layout and not use bcache at all, only use
> Btrfs directly on the whole drives, although I think it's reasonably
> simple to use dmcrypt if needed/desired. But it's still better for
> troubleshooting to make the storage stack as simple as possible.
> Without more debugging information from all the layers, it's hard to
> tell which layer to blame without just using the big stick called
> process of elimination.
>
> Maybe Qu has some ideas based on the call trace though - I can't parse it.
>
> --
> Chris Murphy



[Index of Archives]     [Linux Filesystem Development]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux