xfs
[Top] [All Lists]

XFS weirdness

To: linux-xfs@xxxxxxxxxxx
Subject: XFS weirdness
From: Sean Dougherty <sean@xxxxxxxxxxxxxxxxxxxxx>
Date: Wed, 28 Feb 2001 13:55:20 -0600 (CST)
In-reply-to: <20010228202457.A20859@esparrall.udg.es>
Sender: owner-linux-xfs@xxxxxxxxxxx
This might be an xfs issue or not.  I am not sure.  The following only 
seems to happen when I have files over 2G in size, hence I only see it 
on the XFS partitions.

I am using the cvs as of 2/27/01.  Any partitions are created with

mkfs -t -f -d unwritten=0 -l size=32768b /dev/hdxx
they are mounted with
mount -t -o logbufs=4,logbsize=32768 /dev/hdxx /xfs

these are ide drives with both dma and multi count on (all maxtor drives 
at the moment)

I create a directory under /xfs called test then export /xfs/test as an 
NFS file system to a sun box 5.8 (I have nfs3 compiled in the kernel of 
the xfs box).  Now it gets weird.

>From the sun box I nfs mount xfs-server:/xfs/test to /testxfs that works fine
from the sun box I tar some files to the /testxfs.  All is fine and 
dandy...then at some point, and no it never happens at the same place twice,
the tar dies with a write error. usually rpc timeout.  

>From the sun box I do a df or an ls of /testxfs and after a fashion I get 
another rpc timeout.  But on the xfs-server, if I cd to /xfs/test and try 
to do an ls 9 times out of 10 ls will hang.  Nothing short of rebooting 
the box will bring it back.  Also nfs is also hung (cannot kill rpc.nfsd).
The 10th time though I get from the ls "cannot stat ." also nfs if hung.  
After this reboot I have to do an xfs_repair.  One of the many strange 
things is, I can ls /xfs all day...it is just the /xfs/test that is broken.

Has anyone else seen this?  Does anyone have any ideas.

I have tried with ext2 file systems and they seem to work...but again I 
never move a 2G+ file to them.

I am checking with samba next to see if this is an nfs ism.

Sean
Manager of way too much
TTUHSC At Amarillo
sean@xxxxxxxxxxxxxx


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