Date: Fri, 3 Oct 2014 18:10:16 -0400 (EDT) From: Rick Macklem <rmacklem@uoguelph.ca> To: Garrett Wollman <wollman@csail.mit.edu> Cc: freebsd-fs@freebsd.org Subject: Re: 9.3 NFS client bug? Message-ID: <1265026957.57440085.1412374216862.JavaMail.root@uoguelph.ca> In-Reply-To: <21551.6675.236161.476053@khavrinen.csail.mit.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
Garrett Wollman wrote:
> I've been doing some NFS testing, and found some issues with the NFS
> client in 9.3. This is pretty easily reproducible with bonnie++, but
> I haven't tried to reduce the size of the workload to make it easier
> to test or trace.
>
> ##### NFSv3 mount, 9.3 client from 9.3 server
> # bonnie++ -s 0 -n 1024:65536:0:1023:16384 -u 4294967294 -D
> Using uid:-2, gid:65533.
> Create files in sequential order...done.
> Stat files in sequential order...done.
> Delete files in sequential order...Bonnie: drastic I/O error (rmdir):
> Directory not empty
> Cleaning up test directory after error.
>
> Identical behavior is observed for a 9.3 client talking to a 9.1
> server, and for NFSv4 mounts, but NOT for an Ubuntu 14.04 client
> talking to the 9.3 server using either NFSv3 or NFSv4.
>
A couple of things to try:
1 - I suspect you are using TCP, but you might want to check this.
2 - Try an "oldnfs" mount and see if the old client exhibits the same behaviour.
3 - If you could disable the "Cleaning up test..." after the failure, it
would be nice to check and see if "ls" on the directory shows any files.
Then, unmount/remount it and do the "ls" again.
4 - Capture the packets for wireshark. If you can just capture the "Delete files..."
it would make the packet capture a lot smaller.
rick
ps: There is one difference between the old and new NFS client's readdir code that
I am aware of. It was inherited from OpenBSD and I left it in, but it can be
safely removed for FreeBSD. I have no reason to believe it would cause a problem,
but I can easily generate a patch for you to test, if the "oldnfs" mount doesn't
exhibit the problem.
> I haven't tried using a 9.2 or 9.1 client (yes, I still have some
> machines running 9.1 here) to see if the issue is a recently
> introduced one, nor have I traced the client or server code to see
> what's actually going on.
>
> Has anyone seen this before?
>
> -GAWollman
>
> _______________________________________________
> freebsd-fs@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-fs
> To unsubscribe, send any mail to "freebsd-fs-unsubscribe@freebsd.org"
>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1265026957.57440085.1412374216862.JavaMail.root>