Skip to contentSkip to conversation

Strange NFS/reiserfs errors

DISCUSSION Brian Tinsley 3 participants 3 messages Aug 15, 2003

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hi! I've posted this to the NFS mailing list, but wanted to drop it here as well to see if anyone may provide further insight.

---------------------------
Can anyone elaborate on what may have caused the following error:

Aug 14 22:09:09 clarc02 kernel: nfsd-fh: found a name that I didn't expect: files/19
Aug 14 22:09:40 clarc02 last message repeated 2314 times
Aug 14 22:15:56 clarc02 last message repeated 1088 times
Aug 14 22:17:10 clarc02 last message repeated 138 times
Aug 14 22:19:03 clarc02 last message repeated 138 times
Aug 14 22:20:53 clarc02 last message repeated 2 times
Aug 14 22:21:30 clarc02 last message repeated 73 times

The underlying filesystem is reiserfs and there was a high volume of writing taking place (or trying to) during this time. None of the writes succeeded as the client was returned EEXIST errors. However, the writes were reissued in full shortly after the last occurence and everything automagically started working properly again.

I'm trying to decipher the source code, but it's a bit out of my area of expertise ;-)

Thanks!

#1 lore raw

Hi,

searching the archives of the reiserfs-ML and I guess also a general web search would show that this is fixed in 2.4.21!

Regards,
Bernd

Quoted 25 lines · Brian Tinsley

Hi! I've posted this to the NFS mailing list, but wanted to drop it here as well to see if anyone may provide further insight.

---------------------------
Can anyone elaborate on what may have caused the following error:

Aug 14 22:09:09 clarc02 kernel: nfsd-fh: found a name that I didn't
expect: files/19
Aug 14 22:09:40 clarc02 last message repeated 2314 times
Aug 14 22:15:56 clarc02 last message repeated 1088 times
Aug 14 22:17:10 clarc02 last message repeated 138 times
Aug 14 22:19:03 clarc02 last message repeated 138 times
Aug 14 22:20:53 clarc02 last message repeated 2 times
Aug 14 22:21:30 clarc02 last message repeated 73 times

The underlying filesystem is reiserfs and there was a high volume of writing taking place (or trying to) during this time. None of the writes succeeded as the client was returned EEXIST errors. However, the writes were reissued in full shortly after the last occurence and everything automagically started working properly again.

I'm trying to decipher the source code, but it's a bit out of my area of expertise ;-)

Thanks!

#2 lore raw
In reply to Brian Tinsley

Hello!

Quoted 5 lines · Brian Tinsley

Hi! I've posted this to the NFS mailing list, but wanted to drop it here as well to see if anyone may provide further insight.
---------------------------
Can anyone elaborate on what may have caused the following error:

Aug 14 22:09:09 clarc02 kernel: nfsd-fh: found a name that I didn't 

We believe the reason is iget4() race.
We have a workaround implemented in 2.4.21, so consider upgrading. After you upgrade, please run reiserfsck (take latest reiserfsprogs first. Current reiserfsprogs version is 3.6.11) on all reiserfs volumes you have.

Bye,
Oleg