xfs
[Top] [All Lists]

Re: [Bug 347] XFS clearing the disk

To: bugzilla-daemon@xxxxxxxxxxx
Subject: Re: [Bug 347] XFS clearing the disk
From: Chris Wedgwood <cw@xxxxxxxx>
Date: Wed, 21 Jul 2004 17:11:03 -0700
Cc: xfs-master@xxxxxxxxxxx, christopher.g.dorosky@xxxxxxxx
In-reply-to: <200407212125.i6LLPgml012175@xxxxxxxxxxx>
References: <200407212125.i6LLPgml012175@xxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
On Wed, Jul 21, 2004 at 02:25:42PM -0700, bugzilla-daemon@xxxxxxxxxxx wrote:

> I don't have access to the machine.

so, some machine, not yours, but one you heard about all of a sudden
maybe had something go wrong, files are missing but there are no other
clues...  this is really hard to assist with

> Q1. Where are the logs that were mentioned?

wherever the machine logs kernel messages to, depends on the distro
but usually something like /var/log/kern.log or even /var/log/syslog

you could also try: "dmesg -s 262144 > file" on the machine if it
wasn't rebooted

> Q2. Can you answer the questions about the mounting? It is normally
> safe to access an XFS disk through /dev/sd(xx), right?

depends what you mean, obviously you mound it that way but if it's
mounted you shouldn't access the raw device as it can cause
corruptions or even oopsen

> But df would report the same space occupied as before.

is the machine still up, can you try "lsof +L1" on that machine and
see if there are deleted but still referrence files?

if not, unount the disk and run xfs_repair and see it relocates the
lost stuff to /lost+found

> I will hunt around for some logs on a duplicate machine here, and
> see if I can find any XFS messages. Which log shoudl I look at?

does the duplicate machine have the same problem?  if not, then there
isn't any point

> If I can manage to get a disk to fail here, then I will let you
> know.

this is the *only* error report i've heard like this, i don't want to
say it's not XFS but it's a very strange bug and I can't see how XFS
can loose an entire directory structure w/o spewing many more messages
or something more serious if something was wrong

i really do wonder if the data wasn't deleted at some point,
everything looks like that


  --cw


<Prev in Thread] Current Thread [Next in Thread>