Hello,
i posted a question about a possible ext3fs bug a few month ago.
I see this kernel-msg in the messages:
Oct 8 18:30:00 o5s kernel: EXT3-fs error (device sd(8,10)) in ext3_new_inode:
error 28
Oct 8 18:35:00 o5s kernel: EXT3-fs error (device sd(8,10)) in ext3_new_inode:
error 28
Oct 8 18:40:00 o5s kernel: EXT3-fs error (device sd(8,10)) in ext3_new_inode:
error 28
they repead periodicaly, because a logging file is now corrupted due to
this bug.
I have done successfull an ext3 filesystem check but this error occours
then againa few hour's / day's later affecting a different file...
(Sorry i cant tell if the fsck cleaned the file correctly, because i
deleted the affected file and used a backup. Unfortunately now the arkeia
restore database is affected and i have to rebuild it from the tapes,
which takes some hours to complete...)
At time of my first posting to this group's somebody told this is a known
bug and is fixed since 2.4.18.
I have upgraded from 2.4.17 to 2.4.18 to 2.4.19 and this bug affected from
time to time different files.
This bug is extremly anoying! What if our CVS-Repository is affected...
arkeia backup doesn't work now because of its corrupted database... i have
to rebuild it first...
Any hint to help?
Hardware:
A dual AMD MP 1800+ server with an ICP-Vortex Raid5 Array. Tyan S2460
Mainboard with 2GB Registered ECC Ram.
Software:
RedHat 7.2
Kernel 2.4.19
Recent ACL-Patch and all needed tools
acl-2.0.11-0
libacl-2.0.11-0
acl-devel-2.0.11-0
fileutils-4.1.8acl-65.5
Yes, we (must) use quota and acl together.
Kernel config:
<*> Ext3 journalling file system support
[*] Ext3 extended attributes
[*] Ext3 extended attribute block sharing
[*] Ext3 extended user attributes
[*] Ext3 POSIX Access Control Lists
I havn't yet tested if disabling "extended attribute block
sharing" helps...
But since this time the affected filesystem (/var) didn't use acl and
quota i dont think this is the condition mentioned in th acl known-bugs faq...
where acl+quota+extended user attr is said as problematic...
I post this mail to both (ext3 and acl-list) because i dont know if this
problem is a ext3 only problem or if the acl-patches causes this trouble.
If you need more info please mail.
mfg
--
Jens-Uwe Katolla katolla@otris.de
otris software AG http://www.otris.de
Landgrafenstr. 153 Fon +49 (0)231 95 80 69 -0
D-44139 Dortmund Fax +49 (0)231 95 80 69 -44
Hi, On Wed, Oct 09, 2002 at 11:39:12AM +0200, Jens-Uwe Katolla wrote:> i posted a question about a possible ext3fs bug a few month ago. > > I see this kernel-msg in the messages: > Oct 8 18:30:00 o5s kernel: EXT3-fs error (device sd(8,10)) in ext3_new_inode: error 28 > Oct 8 18:35:00 o5s kernel: EXT3-fs error (device sd(8,10)) in ext3_new_inode: error 28 > Oct 8 18:40:00 o5s kernel: EXT3-fs error (device sd(8,10)) in ext3_new_inode: error 28 > > they repead periodicaly, because a logging file is now corrupted due to > this bug.It's not a corruption, you ran out of inodes. There was a bug in some older versions of ext3 which interpreted this particular error as being more serious than it actually is, so the error was logged, but current ext3 kernels should just return ENOSPC up to the application. Cheers, Stephen