Re: filesystem corrupt - error -117

On 2021/10/26 14:03, Mia wrote:

Hi Qu,

thanks for clarification.
So I should just ignore these errors for now?

Yes, none of them is going to cause any direct problems.

What about these ones, you haven't mentioned:
bad metadata [342605463552, 342605479936) crossing stripe boundary

This is the same, it just means it crosses 64K boundary, which is not
supported for the incoming subpage support (using 4K page size on 64K
page size systems).

Problem with updating is that this is currently still Debian 10 and a
production environment and I don't know when it is possible to upgrade
because of dependencies.

OK, understood the situation now.

Then I can't provide much helper as I'm not familiar with Debian...

If not reproducible so far, I can only recommend for a memtest to rule
out memory bitflip, which could also cause the bug.

Thanks,
Qu

Regards
Mia

------ Originalnachricht ------
Von: "Qu Wenruo" [off-list ref]
An: "Mia" [off-list ref]; "Qu Wenruo" [off-list ref];
linux-btrfs@vger.kernel.org
Gesendet: 26.10.2021 00:45:18
Betreff: Re: filesystem corrupt - error -117

quoted

On 2021/10/26 01:09, Mia wrote:

quoted

Hi Qu,

sorry for the late reply. I tried the btrfs check again with arch
live cd:

root@archiso ~ # uname -a
Linux archiso 5.11.16-arch1-1 #1 SMP PREEMPT Wed, 21 Apr 2021 17:22:13
+0000 x86_64 GNU/Linux
root@archiso ~ # btrfs --version
btrfs-progs v5.14.2

https://gist.github.com/lynara/12dcfff870260b6bc35b9d1137921fc4

OK, so the metadata problem is really there, but it shouldn't affect
your fs right now, unless you want to mount it with 64K page size.

And for the new error (inline file extent too large), it may cause
problems, but under most cases, kernel can handle it without problem.

quoted

I'm still getting many errors.
Sorry I currently don't know what caused this. I suspect it might be
Seafile since I'm now having a currupted library there.

Should I use --repair?

No, --repair won't help in this case.

In fact, your fs is fine, no on-disk metadata problem yet.

For your case, I can only recommend to use newer kernel to have better
sanity check.
Meanwhile I would also recommend to run a memtest to ensure it's not
some memory problem causing the bug.

Thanks,
Qu

quoted

Regards
Mia

------ Originalnachricht ------
Von: "Qu Wenruo" [off-list ref]
An: "Qu Wenruo" [off-list ref]; "Mia" [off-list ref];
linux-btrfs@vger.kernel.org
Gesendet: 25.10.2021 13:18:54
Betreff: Re: filesystem corrupt - error -117

quoted

On 2021/10/25 19:14, Qu Wenruo wrote:

quoted

On 2021/10/25 19:13, Mia wrote:

quoted

Hi Qu,

thanks for your response.
Here the output of btrfs check:
https://gist.github.com/lynara/1c613f7ec9448600f643a59d22c1efb2

Unfortunately it's not full, and it's using an old btrfs-progs
which can
cause false alert.

My bad, gist is folding the output.

It shows no corruption for the extent tree, thus I guess the
transaction
abort has prevented COW from being broken.

quoted

Please use latest btrfs-progs v5.14.2 to re-check.

In that case, a newer btrfs-progs is only going to remove the false
alerts.

Any clue on the workload causing the abort?

For now, I can only recommend to use newer kernel (v5.10+ I guess?) to
see if you can reproduce the problem.

Thanks,
Qu

quoted

Thanks,
Qu

quoted

Thanks,
Mia

------ Originalnachricht ------
Von: "Qu Wenruo" [off-list ref]
An: "Mia" [off-list ref]; linux-btrfs@vger.kernel.org
Gesendet: 25.10.2021 12:55:46
Betreff: Re: filesystem corrupt - error -117

quoted

On 2021/10/25 18:53, Qu Wenruo wrote:

quoted

On 2021/10/25 16:46, Mia wrote:

quoted

Hello,
I need support since my root filesystem just went readonly :(

[641955.981560] BTRFS error (device sda3): tree block 342685007872
owner
7 already locked by pid=8099, extent tree corruption detected

This line explains itself.

Your extent tree is no corrupted, thus it allocated a new tree
block

I missed the "w" for the word "now"...

quoted

which is in fact already hold by other tree.

This means your metadata is no longer protected properly by COW.

"btrfs check" is highly recommended to expose the root cause.

quoted

root@rx1 ~ # btrfs fi show
Label: none  uuid: 21306973-6bf3-4877-9543-633d472dcb46
     Total devices 1 FS bytes used 189.12GiB
     devid    1 size 319.00GiB used 199.08GiB path /dev/sda3

root@rx1 ~ # btrfs fi df /
Data, single: total=194.89GiB, used=187.46GiB
System, single: total=32.00MiB, used=48.00KiB
Metadata, single: total=4.16GiB, used=1.65GiB
GlobalReserve, single: total=380.45MiB, used=0.00B

root@rx1 ~ # btrfs --version
                                                :(
btrfs-progs v4.20.1


root@rx1 ~ # uname -a
Linux rx1 4.19.0-17-amd64 #1 SMP Debian 4.19.194-3 (2021-07-18)
x86_64
GNU/Linux

This is a little old for btrfs, but I don't think that's the cause.

Thanks,
Qu

quoted

Hope someone can help.
Regrads
Mia