stale nfs handles

decker · #1 7th May 2004, 05:08 PM

Hi, since upgrading a few servers and all of our desktop machines from Red Hat to Fedora, my users have started noticing a lot of stale NFS handle errors. I'm not entirely sure what it means to be honest. Anyway, has anyone else experienced that problem, and would they happen to have a solution?

crackers · #2 8th May 2004, 04:53 AM

The only time I've run across this where there were a lot of stale handles was when the server's network card was flaky. ifconfig showed a huge number of dropped packets - replaced the card and haven't had any problems since. So that might be the first place to look...

decker · #3 10th May 2004, 09:42 PM

Yeah well, there aren't any errors as far as I can tell. At least when I run dmesg it never lists any.

I can't say it's a whole lot of stale nfs handles though. In fact I've yet to even see it happen, but a lot of my users have complained about them.

crackers · #4 11th May 2004, 03:32 AM

You won't necessarily see dropped packets with dmesg - you have to look at the stats from ifconfig. Just pointing out one starting point... Another thing you can try is to set the read/write block sizes: rsize=8196,wsize=8196 seems to work well for the servers I have set up.

On my home systems, I actually got real paranoid about this kind of issue and I restart the NFS server every night/morning when I know there's no activity on it (5 AM, if you're curious).