EXT2FS problems & 2.2.19 & 3ware RAID

EXT2FS problems & 2.2.19 & 3ware RAID

[Udo Wolter]

Hi !

I'm hosting a server with 120 GB disk space. This disk space consists of
a RAID10 via a 3ware Escalade 6800 controller. As you can see, the disks are
IDE disks (4 x 60 GB) but the controller maps it as scsi-devices to
the system. It runs fine but from time to time I get those errors:

kernel: EXT2-fs warning (device sd(8,6)): ext2_free_inode: bit already cleared for inode 885312
kernel: EXT2-fs warning (device sd(8,6)): ext2_free_inode: bit already cleared for inode 885258

etc.

The messages are coming almost once a week. As long as those messages are not
harming the filesystem, I had no problems with them (in fact, they say, they
are warnings and no errors). But the last times I tried to do a du over the
full partition, I got 2 or 3 files and directories which say that that they
can't be accessed -> I/O-Error. Neither I can't delete them nor I can
access them in any way, not even list.

After using debugfs (read only mode during the mount) I can see the files
in the listing but even here they are not accessible.

I shut the machine down for 2 hours to do a fsck and the errors had been
gone but after some days the errors came back (different files and directories,
not the same). At this time it's not possible to shut it down because
more than 5000 users use it at the moment. Anyway before doing a fsck again
I'd like to solve the problem so that the system couldn't get corrupted again.

The kernel-version is 2.2.19. Maybe switching to 2.4.x helps but I really
like to know what has happened and why, before I'll do such a big step.

BTW, 3ware has a tool to check for the disks etc. This tool says that
everything is fine so it can only be an EXT2FS problem.

If you'd like to answer my questions, please do it via Mail-CC because I don't
read all of this mailing list (only some parts of the archives, this is heavy
enough for me :))

Thanx for any hints !

Bye,
	Udo

-- 
Udo Wolter
email:    uwp@dicke-aersche.de
www:      www.dicke-aersche.de
Fonautic:   0700 - UDO - 00000 

Re: EXT2FS problems & 2.2.19 & 3ware RAID

[Luigi Genoni]

On Thu, 14 Jun 2001, Udo Wolter wrote:

> Hi !
>
> I'm hosting a server with 120 GB disk space. This disk space consists of
> a RAID10 via a 3ware Escalade 6800 controller. As you can see, the disks are
> IDE disks (4 x 60 GB) but the controller maps it as scsi-devices to
> the system. It runs fine but from time to time I get those errors:
>
> kernel: EXT2-fs warning (device sd(8,6)): ext2_free_inode: bit already cleared for inode 885312
> kernel: EXT2-fs warning (device sd(8,6)): ext2_free_inode: bit already cleared for inode 885258
>
when i saw something similar, with a configuration really similar to your,
but on scsi disks, it was because a disc was near to break.
I had to change the disk and everything went ok.

It also happened to me after a crash on a disk due to eathing
problems, last august. Same messages,
and thanx God the disk was not really damnaged, so low level formatted
the disk, rebuilt partition table, then I rebuilt the fs to restore my
backup and everything went ok.
So my suggestion is to loock for HW problems.

> etc.
>
> The messages are coming almost once a week. As long as those messages are not
> harming the filesystem, I had no problems with them (in fact, they say, they
> are warnings and no errors). But the last times I tried to do a du over the
> full partition, I got 2 or 3 files and directories which say that that they
> can't be accessed -> I/O-Error. Neither I can't delete them nor I can
> access them in any way, not even list.
>
> After using debugfs (read only mode during the mount) I can see the files
> in the listing but even here they are not accessible.
>
> I shut the machine down for 2 hours to do a fsck and the errors had been
> gone but after some days the errors came back (different files and directories,
> not the same). At this time it's not possible to shut it down because
> more than 5000 users use it at the moment. Anyway before doing a fsck again
> I'd like to solve the problem so that the system couldn't get corrupted again.
>
exactly, that what I saw myself on the soon to get broken disk.

bests

Luigi Genoni