Help with NFS reliability problem

Hi,

I am having reliability issues with a NFS mount.

I have a directory hierarchy containing several thousands files and
directories that get exported from a server to a client.

The only options I am using in the server /etc/exports are:
(rw,no_root_squash)

I realise this is a potential security risk but this is a private
network with no external access so at the moment, I am happy with
that and am more concerned with the reliability issues.

I used no special option apart from rw in the client fstab

Operating system:

CentOS 4.1
rpm -q nfs-utils => nfs-utils-1.0.6-70.EL4


My problem:

Occasionally I am unable to access one of the directory nor its
content. On investigation, doing a "ls -l" gives me the following:

drwxr-xr-x 4 root root 4096 Jan 31 2007 .
drwxr-xr-x 30 root root 4096 Jan 31 2007 ..
drwxr-xr-x 6 root root 4096 Jan 31 2007 37705
?--------- ? ? ? ? ? 45644

when I would have expected a normal listing to look like:

drwxr-xr-x 4 root root 4096 Jan 31 2007 .
drwxr-xr-x 30 root root 4096 Jan 31 2007 ..
drwxr-xr-x 6 root root 4096 Jan 31 2007 37705
drwxr-xr-x 6 root root 4096 Jan 31 2007 45644

Does anybody know why this is happening and how to fix it?

I've read various NFS guides. Didn't find a useful answer. I am
considering remounting using both "tcp" and "sync" option and
potentially "wno_delay" in the hope that this would solve my
reliability issue but doing so would be a blind shot in the dark
crossing my fingers and hoping. An explanation of the cause of the
problem and how to fix it would be a lot more acceptable.

The added problems is that I don't have the liberty to try all
possible option combinations until I find something that work and the
problem happens randomly without known way to reproduce it. It has
happened twice in 5 days, was temporary solved with umounting and
remounting the client.

Any help would be greatly appreciated.

Thanks

Yan

Yannick Tremblay wrote:
> Hi,
>
> I am having reliability issues with a NFS mount.
>
> I have a directory hierarchy containing several thousands files and
> directories that get exported from a server to a client.
>
> The only options I am using in the server /etc/exports are:
> (rw,no_root_squash)
>
> I realise this is a potential security risk but this is a private
> network with no external access so at the moment, I am happy with
> that and am more concerned with the reliability issues.
>
> I used no special option apart from rw in the client fstab
>
> Operating system:
>
> CentOS 4.1
> rpm -q nfs-utils => nfs-utils-1.0.6-70.EL4
>
>
> My problem:
>
> Occasionally I am unable to access one of the directory nor its
> content. On investigation, doing a "ls -l" gives me the following:
>
> drwxr-xr-x 4 root root 4096 Jan 31 2007 .
> drwxr-xr-x 30 root root 4096 Jan 31 2007 ..
> drwxr-xr-x 6 root root 4096 Jan 31 2007 37705
> ?--------- ? ? ? ? ? 45644
>
> when I would have expected a normal listing to look like:
>
> drwxr-xr-x 4 root root 4096 Jan 31 2007 .
> drwxr-xr-x 30 root root 4096 Jan 31 2007 ..
> drwxr-xr-x 6 root root 4096 Jan 31 2007 37705
> drwxr-xr-x 6 root root 4096 Jan 31 2007 45644
>
> Does anybody know why this is happening and how to fix it?
>
> I've read various NFS guides. Didn't find a useful answer. I am
> considering remounting using both "tcp" and "sync" option and
> potentially "wno_delay" in the hope that this would solve my
> reliability issue but doing so would be a blind shot in the dark
> crossing my fingers and hoping. An explanation of the cause of the
> problem and how to fix it would be a lot more acceptable.
>
> The added problems is that I don't have the liberty to try all
> possible option combinations until I find something that work and the
> problem happens randomly without known way to reproduce it. It has
> happened twice in 5 days, was temporary solved with umounting and
> remounting the client.

Extra options shouldn't be needed. My NFS servers are up to
354days of uptime... no issues (not longer simply because we
usually have one mandatory shutdown during a year).

I'm running SUSE 9.3 on my servers currently (just NFSv3, and
we have a few "at risk" accessors running udp instead of
tcp (because of the age of the Unix OS hitting them).

I'm going to guess that your problem could be network
related, but there could be a bug in NFS that is in the
newer code.... not sure.

All I'm saying is that NFS is very reliable in Linux.
We house hundreds of gigabytes of home directories and
tons of software development build areas....

>
> Any help would be greatly appreciated.
>
> Thanks
>
> Yan

In comp.os.linux.networking Yannick Tremblay <(E-Mail Removed)>:

> I am having reliability issues with a NFS mount.
[..]

> CentOS 4.1
> rpm -q nfs-utils => nfs-utils-1.0.6-70.EL4

I'd install all patches including kernel there should be tons,
reboot and check if this improves things, preferable using 'yum
update'.

Good luck

[..]

--
Michael Heiming (X-PGP-Sig > GPG-Key ID: EDD27B94)
mail: echo (E-Mail Removed) | perl -pe 'y/a-z/n-za-m/'
#bofh excuse 63: not properly grounded, please bury computer