Hi all,

I was hoping for a couple of neat tips that would help me get our NFS usage down a bit when multiple users log into our Linux cluster. At the moment we offer KDE by default as it's usually pretty stable and intuitive to use, with each user's /home directory mounted via NFS so that they get the same profile on every machine. The problem comes when around 10 or more users sit down at our cluster simultaneously, as the NFS usage starts to climb, response time for running programs increases and people start to mutter in my direction.

I already redirected the cache from Firefox to /tmp and installed as many programs as possible locally on each machine, but it'd be great if I could figure out how to slash the NFS usage of normal KDE processes as well? Would copying the .kde4 folder to somewhere local during login, then replacing it by a symlink in /home help at all, or is it likely to be services such as strigi, Akonadi etc that are indexing the NFS-mounted /home too often? If so, can I easily exclude files on the NFS mount from indexing, and are there any files that I need to keep indexing?!

I found a tip that suggested changing the flags for the NFS mount to 'async', but that sounds a bit risky on an open cluster where crashes occur, power cables get pulled out and users have a habit of hard-rebooting first and asking questions later. Especially as *some* of the data stored there is actually important.

All tips appreciated!!

You can probably just disable Strigi entirely, if nobody is using its functionality. It's not really required, unless you want users to be able to search for files by content.

Ok, thanks! I wasn't sure about all the KDE programs that seem to be dependent on strigi, I didn't want to switch it off and risk breaking KDE. So long as it's not essential then I guess I can give it a go.