Tags:
Apple Computer, WorkGroup Manager, Leaopard Server
A couple months ago we were having intermittent problems with AFP on our dual CPU G5 Xserve running Tiger. At a certain level of usage (maybe 75 concurrent network logins), the AppleFileServer process would hit 200% CPU usage and all users would "beachball." Sometimes we could just wait it out, other we'd have to restart the AFP process. We did some server tuning according to a Apple Technical document and an AFP548.com article in hopes alleviate the problem and it seemed to help, but still happened. We didn't think too much of it at the time because it would only happen every couple weeks or so and we planned on getting a new Intel Xserve anyway.
Well, we have the new Xserve running 10.5.1 and the same problem (or a very similar one) just started again today. At some point the AppleFileServer process will just start hogging CPU (380%) and sometimes RAM, and nobody can connect or do anything at all until I kill the process and restart it. The first time it happened today I actually had to reboot the whole server because even Server Admin was unable to connect. I couldn't monitor the AFP service to see how many people were connected and such.