xfs
[Top] [All Lists]

Re: Strange df output

To: Eric Sandeen <sandeen@xxxxxxx>
Subject: Re: Strange df output
From: Jarrod Johnson <jbj-ksylph@xxxxxxxxxxxxxxxx>
Date: Sat, 17 Jan 2004 09:16:14 -0500
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <1073595887.27384.260.camel@stout.americas.sgi.com>
References: <20040108154940.7dd28591.jbj-ksylph@ken-ohki> <1073595887.27384.260.camel@stout.americas.sgi.com>
Sender: linux-xfs-bounce@xxxxxxxxxxx
Unfortunately, this was first created ~1.5 years ago, so I have no mkfs 
arguments, I assume the default from whatever it was back then.  I've gone 
since to kernel 2.4.23 with the snapshot for download applied, and it was about 
then I started noticing this behavior.  The system crashed at one point listing 
about ~4G free, but on reboot the system had 0 free.

I tried to run xfs_check on it, but xfs_check is killed partway through, kernel 
prints:
__alloc_pages: 0-order allocation failed (gfp=0x1d2/0)
VM: killing process xfs_db

Is there some serious issues going on here?  Amidst 344G of data, it's hard to 
find if only a portion of data is corrupt.  By the same token, I don't have any 
storage on which to back up 344GB of filesystem for a new mkfs.

Anyone have any suggested path to ensure things are ok, or fix things if they 
aren't?  

The only other interesting aspect that may be of interest is that the 
filesystem is on a software RAID-5 volume.

On 08 Jan 2004 15:04:47 -0600
Eric Sandeen <sandeen@xxxxxxx> wrote:

> Any chance you can give us a reproducible way (from mkfs) to demonstrate
> the problem?
> 
> You could be seeing some effects of preallocated space being freed (esp.
> as the fs fills up), other than that I'm not certain yet.
> 
> -Eric
> 
> 
> On Thu, 2004-01-08 at 14:49, Jarrod Johnson wrote:
> > I'm running xfs on a software raid5 array, the kernel version is:
> > SGI XFS snapshot-2.4.23-2003-12-01_00:33_UTC with ACLs, no debug enabled
> > 
> > And here is a taste of what I'm seeing:
> > df -k .:
> > Filesystem           1K-blocks      Used Available Use% Mounted on
> > /dev/md0             360138240 356276448   3861792  99% /storage
> > dd if=/dev/zero of=tst bs=1000000 count=300
> > df -k .: 
> > Filesystem           1K-blocks      Used Available Use% Mounted on
> > /dev/md0             360138240 356327136   3811104  99% /storage
> > rm -f tst:
> > Filesystem           1K-blocks      Used Available Use% Mounted on
> > /dev/md0             360138240 355983080   4155160  99% /storage
> > 
> > So I had some space free, making a 300,000,000 byte file only consumed 
> > about 5,000k, and then deleting it freed up more space than I had to start 
> > with?
> > 
> > This is consistantly occuring.  A week or so back, it inexplicably went 
> > from ~2GB free to full instantaneously.  I didn't believe the person was 
> > keeping good track of it and just filled it up, so I did this a few times 
> > and if this is happening, I now believe that what the person said is true.
> > 
> > Why is the free space not decreasing on creation of new files, yet 
> > increasing on deletion?
> > 
> > Please CC me as I'm not subscribed.
> -- 
> Eric Sandeen      [C]XFS for Linux   http://oss.sgi.com/projects/xfs
> sandeen@xxxxxxx   SGI, Inc.          651-683-3102
> 



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