NFS latencies and strange logs when using NFSv3

[NE]

Hi,

I'm using FreeBSD 9.1-RELEASE with an NFS exposed filesystem for storage, with only NFSv2 and NFSv3 available:

$ rpcinfo -u filer.darkstar.net nfs
program 100003 version 2 ready and waiting
program 100003 version 3 ready and waiting
$

The filesystem is mounted using the following fstab entry:

filer.darkstar.net:/users/ping1/unix /home/ping1  nfs  rw,hard,bg,intr,retrycnt=2 0 0

Finally, the following NFS related stuff has been added into the /etc/rc.conf:

rpc_lockd_enable="YES"
rpc_statd_enable="YES"
nfs_client_enable="YES"

With this setup I am experiencing latencies while not facing them on the Linux workstations (with mount options hard,bg,intr,retry=2). Furthermore, kernel logs are reported problems in a multiplexed way, as in the following example:

(10:17:37) ping1: Sep 20 09:20:13 plonk kernel: newnfs server filer.darkstar.net:/users/ping1/unix: is alive again
(10:17:37) ping1: Sep 20 09:20:13 plonk kernel: newnfs server filer.darkstar.net:/users/ping1/unix: ins eawlnifvs e saegraivne

Any idea?

Thanks,

:Nicolas

 

[NE]

After reading some unrelated but very interesting threads of filestystem tuning, I realize FreeBSD mounts are done using the noasync option by default while Linux ones are using async. I will investigate this.

 

[NE]

After investigations, still no solution found. Using both sync or noasync has no effect and I still have long locks.