Poor NFS performance in Ubuntu Breezy with fs mounted from a Debian 3.1 server

[Al Gordon]

On 4/25/06, Erling Ringen Elvsrud <systemansvarlig at vagaungdomsskule.no> wrote:
> Hello,
>
> I'm currently working with a authentication / file storage setup for a
> highschool (small one, about 150 students).
>
> LDAP works pretty well so far, but NFS is the current problem.
> I export a filesystem from a P3 1400 MHz / 512MB / 73GB SCSI server
> running Debian 3.1 (Stable). I export it with a /etc/exports file
> looking like this:
>
> /heim1     192.168.1.0/24(rw,sync,no_root_squash)
>
> (I'm aware that no_root_squash is a security problem, but this is not a
> production system, I'm just testing).
>
> The clients then mount it on /heim.
> (tor.intern.vagaungdomsskule.no:/heim1    /heim  nfs
> rsize=4096,wsize=4096,rw,suid,dev,auto,nouser,sync    0       0)
>
> The problem is: When I mount the filesystem from my workstation running
> Ubuntu Breezy (with the fstab entry above) I get very poor performance.
> I perform a very simple test in a directory on the nfs mounted
> filesystem from my workstation:
>
> time dd if=/dev/zero of=testfile bs=16k count=1000
> 64+0 records in
> 63+0 records out
> 1032192 bytes transferred in 6,086174 seconds (169596 bytes/sec)
>
> real    0m6.140s
> user    0m0.001s
> sys     0m0.014s
>
> Something is clearly wrong. When I mount the same nfs exported filsystem
> on a testserver running Debian 3.1 I get much better performance
> (somewhere around 10MB/Sec, which is a good starting point).
>
> In the log of the server (/etc/syslog) I get some messages like this
> (192.168.1.60 is the Workstation running Ubuntu):
>
> Apr 25 11:49:30 localhost kernel: lockd: cannot monitor 192.168.1.60
> Apr 25 11:49:30 localhost kernel: nsm_mon_unmon: rpc failed, status=-13
> Apr 25 11:49:30 localhost kernel: lockd: cannot monitor 192.168.1.60
> Apr 25 11:49:31 localhost kernel: nsm_mon_unmon: rpc failed, status=-13
> Apr 25 11:49:31 localhost kernel: lockd: cannot monitor 192.168.1.60
> Apr 25 11:49:31 localhost kernel: nsm_mon_unmon: rpc failed, status=-13
> Apr 25 11:49:31 localhost kernel: lockd: cannot monitor 192.168.1.60
> Apr 25 11:49:32 localhost kernel: nsm_mon_unmon: rpc failed, status=-13
> Apr 25 11:49:32 localhost kernel: lockd: cannot monitor 192.168.1.60
>
> I suspect this have something to do with the poor performance. Som
> googling suggested that lockd was not running on the client, but it
> seems to do. I also added NEED_LOCKD=yes in /etc/default/nfs-common just
> in case.

Do you have portmap installed/running on both the server and the client side?

--

  -- AL --