Multiple OpenAFS Problems

[fireboy1919]

I'm using OpenAFS with the 1.3.2-r1 ebuild, and I've been experiencing a lot of problems with the filesystem:
1) Attempting to delete a large file results in the fileserver process thrashing, and it must be killed with kill -9 in order to restart.
2) Filesystem access uses a lot of CPU time.
3) Occasionally, the filesystem will become unresponsive for no apparent reason. Later, it will become active again. The logs do not indicate a reason for this unless the Salvager is running (which is not always the case).
4) After using the salvager, the system will often not respond at all for about an hour (filesystem access is not possible).
5) Attempting to remove the AFS kernel module results in a kernel panic. The /afs mountpoint cannot be unmounted. AFSD cannot be stopped or killed, and restart seems to be impossible.

A few of these problems are very serious (most noteably #1, #3, and #4), and would lead me to classify this as Alpha-level software. Can anyone suggest ways around these problems? Does anyone have any information about why these things happen?

[morgant]

While I consider afs to be a pretty ornery piece of software, 1.3.2 is particularly painful. Of course, 1.2.9 is the latest stable release, not 1.3.2. I tried working through all the difficulties in 1.3.2, but it became much too difficult and didn't have anywhere near the server stability I was used to (such as in 1.2.8). In essence, what I'm recommending is using 1.2.9 unless you're interested in hacking the afs code. It's not in the portage tree, but it's pretty easy to modify the 1.3.2 ebuild to install 1.2.9 instead. If for some reason you don't want to step back, I'd recommend trying the openafs mailing lists for help with getting these problems resolved.
_________________
Morgant

"Computers are like Old Testament gods; lots of rules and no mercy."