xfs_repair keeps reporting errors

xfs_repair keeps reporting errors

[Emmanuel Florac]

[-- Attachment #1: Type: text/plain, Size: 1587 bytes --]

Hi,

I just had a problem with a RAID array, now that the rebuild process
is complete, as I run xfs_repair (v 5.0) again and again it keeps
reporting problems (here running xfs_repair for the 3rd time in a row):


bad CRC for inode 861144976062
bad magic number 0x0 on inode 861144976062
bad magic number 0x0 on inode 217316006460
bad version number 0x0 on inode 217316006460
inode identifier 0 mismatch on inode 217316006460
bad version number 0x0 on inode 861144976062
bad CRC for inode 217316006461
bad magic number 0x0 on inode 217316006461
inode identifier 0 mismatch on inode 861144976062
bad version number 0x0 on inode 217316006461
inode identifier 0 mismatch on inode 217316006461
bad CRC for inode 217316006462
bad magic number 0x0 on inode 217316006462
bad CRC for inode 861144976063
bad magic number 0x0 on inode 861144976063
bad version number 0x0 on inode 861144976063
bad version number 0x0 on inode 217316006462
inode identifier 0 mismatch on inode 217316006462
bad CRC for inode 217316006463
bad magic number 0x0 on inode 217316006463
inode identifier 0 mismatch on inode 861144976063
bad version number 0x0 on inode 217316006463
inode identifier 0 mismatch on inode 217316006463

Is there anything else to do?

-- 
------------------------------------------------------------------------
Emmanuel Florac     |   Direction technique
                    |   Intellique
                    |	<eflorac@intellique.com>
                    |   +33 1 78 94 84 02
------------------------------------------------------------------------

[-- Attachment #2: Signature digitale OpenPGP --]
[-- Type: application/pgp-signature, Size: 163 bytes --]

Re: xfs_repair keeps reporting errors

[Brian Foster]

On Thu, Oct 31, 2019 at 03:40:49PM +0100, Emmanuel Florac wrote:
> Hi,
> 
> I just had a problem with a RAID array, now that the rebuild process
> is complete, as I run xfs_repair (v 5.0) again and again it keeps
> reporting problems (here running xfs_repair for the 3rd time in a row):
> 

Problems with the same inodes or the same general issue with different
inodes after repeated repair runs?

What kind of RAID event was involved? Was the filesystem known healthy
prior to the event?

> 
> bad CRC for inode 861144976062
> bad magic number 0x0 on inode 861144976062
> bad magic number 0x0 on inode 217316006460
> bad version number 0x0 on inode 217316006460
> inode identifier 0 mismatch on inode 217316006460
> bad version number 0x0 on inode 861144976062
> bad CRC for inode 217316006461
> bad magic number 0x0 on inode 217316006461
> inode identifier 0 mismatch on inode 861144976062
> bad version number 0x0 on inode 217316006461
> inode identifier 0 mismatch on inode 217316006461
> bad CRC for inode 217316006462
> bad magic number 0x0 on inode 217316006462
> bad CRC for inode 861144976063
> bad magic number 0x0 on inode 861144976063
> bad version number 0x0 on inode 861144976063
> bad version number 0x0 on inode 217316006462
> inode identifier 0 mismatch on inode 217316006462
> bad CRC for inode 217316006463
> bad magic number 0x0 on inode 217316006463
> inode identifier 0 mismatch on inode 861144976063
> bad version number 0x0 on inode 217316006463
> inode identifier 0 mismatch on inode 217316006463
> 
> Is there anything else to do?
> 

I think it's hard to say what might be going on here without some view
into the state of the fs. Perhaps some large chunk of the fs has been
zeroed given all of the zeroed out on-disk inode fields? We'd probably
want to see a metadump of the fs to get a closer look.

Brian

> -- 
> ------------------------------------------------------------------------
> Emmanuel Florac     |   Direction technique
>                     |   Intellique
>                     |	<eflorac@intellique.com>
>                     |   +33 1 78 94 84 02
> ------------------------------------------------------------------------

Re: xfs_repair keeps reporting errors

[Emmanuel Florac]

[-- Attachment #1: Type: text/plain, Size: 838 bytes --]

Le Fri, 1 Nov 2019 06:45:51 -0400
Brian Foster <bfoster@redhat.com> écrivait:

> I think it's hard to say what might be going on here without some view
> into the state of the fs. Perhaps some large chunk of the fs has been
> zeroed given all of the zeroed out on-disk inode fields? We'd probably
> want to see a metadump of the fs to get a closer look.
> 

Alas, the metadump is 28 GB... (this is a 500 TB filesystem). So far it
seems to behave better with no background RAID controller operation.

-- 
------------------------------------------------------------------------
Emmanuel Florac     |   Direction technique
                    |   Intellique
                    |	<eflorac@intellique.com>
                    |   +33 1 78 94 84 02
------------------------------------------------------------------------

[-- Attachment #2: Signature digitale OpenPGP --]
[-- Type: application/pgp-signature, Size: 163 bytes --]

Re: xfs_repair keeps reporting errors

[Eric Sandeen]

[-- Attachment #1.1: Type: text/plain, Size: 626 bytes --]

On 11/4/19 7:31 AM, Emmanuel Florac wrote:
> Le Fri, 1 Nov 2019 06:45:51 -0400
> Brian Foster <bfoster@redhat.com> écrivait:
> 
>> I think it's hard to say what might be going on here without some view
>> into the state of the fs. Perhaps some large chunk of the fs has been
>> zeroed given all of the zeroed out on-disk inode fields? We'd probably
>> want to see a metadump of the fs to get a closer look.
>>
> 
> Alas, the metadump is 28 GB... (this is a 500 TB filesystem). So far it
> seems to behave better with no background RAID controller operation.

They sometimes compress quite well FWIW.

-Eric


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 873 bytes --]