afp server issue - very hign cpu load

I had one dual quad server at over 600% yesterday. People complaining of slowness and spinning wheel. The 2009 001 update doesn't seem to have fixed the issue. My ticket has been reopened with apple

Just wanted to add that I have been seeing the same problem and no, the security update on the server did not fix the issue. From my Apple SE's recommendation I am installing the security update to all workstations as well. Don't know if anyone has done that yet or has any results? I am at a school district and have 450 computers to update and over half are still 10.5.5, so this will take a while.

Just my 2 cents. Not resolved either. After all the modifications to AppleShareClient, turning off caching on the xRAID, things were definitely much better. No spikes to 100%, mostly dropping back to low levels consistently, no complaints from users (the most important measure). So a week.
Since applying the patch to the server and all clients: Inconsistent. First Monday to 100% again. Killed the PID, no issues at all until Wednesday. Killed again. Thursday and Friday, instead of the usual abrupt rise in CPU, a gradual increase over the course of the day. Monday of the next week nothing. Tuesday spiked to 100%. Wednesday nothing. Thursday abrupt spike. Friday gradual increase. This week back to the usual rapid rises. The only thing that's consistent since the initial mods thru the patch: only 2 spikes to 100% and users don't seem to be beachballing.
So for now, killing the PID every night. Keeping my fingers crossed for something better.
Also - I'd asked earlier about the v1.1 upgrade that came out before the Security patch - any one able to apply it? (I could on my SUS server, doesn't show up on the AFP server.) Did it do anything?

Replying to myself. Sorry.

AFP 10.5.6 (9G55)

after v1.1

SUS 10.5.6 (9G71)

Anyone have 9G71 on your AFP server?

Both of mine are 9G55

It's amazing that this problem has gone this long!!! This is an AFP problem which apple is responsible for creating. Just like everyone else, since applying the 10.5.6 update on the server over Christmas break, the server CPU spikes to 798% CPU and all the clients crash right when logging out from the Open Directory. I re-imaged 85 clients back to 10.5.5 to temporarily fix the issue. I've turned off Spotlight indexing on the server, no fix. I've deleted the Microsoft user folder, the users' fonts folder and many other tweaks that have been brought up to fixing this issue and still doesn't solve any problems. There is no doubt that it starts when multiple users first login to the system. The problem goes away after they all logoff. I've thought about switching the user folder to NFS only problem is that the files they have created aren't recognized properly after logging back in using NFS. This would be a complete headache to to solve for 400 plus students. Has anyone reinstalled the server back to 10.5.4 to see if the AFP service problems goes away. I never had an issue until applying this update and the new security patch did absolutely nothing to fix it.

We're seeing this too... the server returning to normal when everyone goes home.

I hope someone with this problem goes as far as to rollback to 10.5.4 to let us know if the problem goes away.

We need this fixed in 10.5.7 and no later.

My only concern on rolling the server back is if there is something stored in the users preferences after the 10.5.6 update that is causing this behavior on the server. If I roll the server back, do I need to clear out the users preferences? Who knows, either way, it sounds like a job that will take through out the night to do, only to hope it will even work!

I don't think rolling back to 10.5.4 will fix your problem as the original poster was on 10.5.4.
I also think I was on 10.5.4 in October when I started getting this problem, and am glad to have found this thread as it proves that its a flaw with the software and not a problem with how I set up our server!

Although some recent driver maintenance on our Raid Card and a couple of tricks here and there seemed to keep us from beach balling, we were stuck at about 80-95% constant processor load. Restarting would temporarily bring it back down but it would build up again very quickly.

After reading about the security update here, I decided to give it a shot. After 3 days of regular use it seems to be holding roughly at 10-20% load without any major spikes or pinning. See the image below.

I'm thinking possibly the reason this is working for me is that I only have 10 users while many posters here easily have 50 or more. Also, I have one common user for all rather than multiple users and that seems to be helping with permissions problems that I was experiencing as well after the constant crashing.

Will give a processor load update here next week (crossing fingers).

!http://djspinz.com/images/server_pic.jpg!

It's funny you should mention the RAID card... although we're seeing a server without the RAID card being hit too. We first wondered if the RAID card played a role somehow.

Our impacted server(s) can have around 300 simultaneous users. About 40-50% of those are AFP.

I have XServes with and without raid cards on 10.5.6 and I still get the problem with the CPU maxing out with any number of users 10-50. My only work around has been to to kill the afp server process and let it re-spawn. Doing that brings it back to a more "normal" mode of operation.

Well after many trials and errors, I decided to start looking at other solutions such as more bandwidth. So I decided to check the status of my link aggregated ports and found out that after applying the 10.5.6 update, it degraded one of the links! So I made sure that all the network cards were properly connected at a gig to the switch. I fixed the link aggregations and actually decided to pair 3 ports instead of only two. My cpu still goes up to 780%, however it does goes back down to normal after my users all login and the network home folders seem to be back to normal and my users are noticing a dramatic increase in speed. So my advice is to other users, check your ports, especially if you have any aggregated to make sure the updates and patches didn't kill it.