[clug] NFS problem?

[Eyal Lebedinsky]

I recently started to see unusual behaviour accessing files over nfs.

I have a server (e7) which collects data, each item is one line in a file.
There are about a dozen such files in a directory. This fs is nfs exported.

I can watch the data arriving with something like
	tail -n 1 -f /path/*.log
and everything looks good.

On my workstation (e4) I run the same command
	tail -n 1 -f /nfs-path/*.log
I now see many updates show as an empty line. If I look at such a file
(less or vi) I see a bunch of zeroes at the end of the file. The actual
file (or e7) does not have those zeroes.

Furthermore, after some time (up to 10 minutes) the data starts to look
OK for a while, then goes bad again.

Another observation: 'ls -l' shows the same (correct) file size on both
systems, and I can see that as data is appended to a file there are more
zeroes presented. Then after a few minutes all the data in the file
is shown correctly.

BTW, if I umount/mount the relevant fs then the data looks good for a
short while before going off the rails as described above.

I noticed this only very recently. The server had no updates during this
period (it is an old f19) but the workstation is following f24.

It is as if nfs is very slow to refresh the exported data.

Did anyone see something similar? Anyone has an idea?

-- 
Eyal Lebedinsky (eyal at eyal.emu.id.au)