NFS I/O or permission denied errors

[Rick Macklem]

Attila Bogar wrote:
> On 03/08/12 01:03, Attila Bogár wrote:
> > See the attached bash script.
> Sorry, my script has been stripped off by the list.
> 
> It's nothing rocket science, takes an export list from a file
> (/root/fs.txt) and mounts the nfs shares in a loop.
> 
> -- 8< --
> #!/bin/bash
> 
> sleepmount=0.33
> 
> function mount.all() {
> set -e
> j=0
> while read fs
> do
> echo mounting $fs
> sleep $sleepmount
> mount -t nfs -v -o sec=sys,vers=3,proto=tcp server:$fs /mnt/$j
> sleep $sleepmount
> umount -f /mnt/$j

Oops, I didn't see that the client was a Linux one. I have no idea
what the semantics of a Linux "umount -f" is, although I'd still
try it without "-f" I was you.

> j=$[j+1]
> done < /root/fs.txt
> set +e
> }
> 
> function umount.all() {
> umount -f /mnt/* &>/dev/null
> }
> 
> fsno="$(wc -l /root/fs.txt|cut -d' ' -f1)"
> umount.all
> i=0;while [ $i -lt $fsno ]; do echo $i; mkdir /mnt/$i ; i=$[i+1]; done
> 
> i=0
> while [ $i -lt 1000 ]
> do
> i=$[i+1]
> echo "RUN: $i" | tee /tmp/nfstest.status
> mount.all
> done
> -- 8< --
> 
> sec=sys,vers=3,proto=udp seems stable, tcp gets an I/O error within
> 2-3
> minutes.
> 
> I'm testing this from an up to date CentOS 6.3 nfs client.
> 
> I suspicious about the network stack. I tried turning off msi and TSO
> on
> the bce card, but no change yet.
> I also took down my lagg interface and tried the onboard single bce,
> no
> change.
> 
> This gentleman had a similar problem with nfsv4 but he got no hint
> from
> the list:
> http://lists.freebsd.org/pipermail/freebsd-stable/2012-June/068658.html
> 
> Can someone reproduce the problem or it is specific to my hardware?
> 
> Thanks,
> Attila
> 
> _______________________________________________
> freebsd-fs at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-fs
> To unsubscribe, send any mail to "freebsd-fs-unsubscribe at freebsd.org"