Comment 73 for bug 561210

Just in case this can help someone (if not sorry for my frequent spam):

After long time of investigation, which clients really have that look-ups and which not i concluded that in my case the problem was caused by the simple fact, that those clients where solely NATed (most of them VMs, Ubuntu 9.04 .. 10.10). And correct I should have realized earlier that NFS over NAT is a 'not so good' idea (bidirectional port handling, file locks etc.). But the resulting errors are very similar to what is described in the report here.

So what can help, if getting a dedicated IP for all machines is not an option?

Do not mount the shares via fstab, instead take a look at autofs, a really great utility for doing on demand mounts and also handling not 100% reliable connections.

Sample config after apt-get install autofs5:

add to /etc/auto.master:
/mnt/myserver /etc/auto.guests --timeout=60 --ghost

create /etc/auto.guests and add (modify defaults to your demands):
/guests -defaults myserver:/your_export_path/guests

create a symbolic link where you really want to access the share:
ln -s /home/guests /mnt/myserver/guests

One thing that remains - now i can write huge files to the NFS server even from clients behind a NAT but the speed decreases the larger the file is (which remembers me on #71 - but i can't find a similar mount option in ubuntu). Anyone any idea?