ProLiant Servers (ML,DL,SL)
1753828 Members
8355 Online
108806 Solutions
New Discussion юеВ

Re: High CPU Utilization

 
Ayman Altounji
Valued Contributor

High CPU Utilization

We have 3 Compaq DL380's running Netware 5.1 and each server is experiencing intermittent high CPU utilization. These are out of the box installs, nothing special. Once the utilization spikes, it doesn't return to normal cpu% util after a reboot. After which the system will function normally (4-10% Cpu) for approx 2 days before cpu% spikes to 60-70% again, therefore requiring another reboot. I have unloaded all un-necessary NLM's and still have the same problem, there are no indications from Netware monitor what is maxing out the CPU, could the problem be hardware related?
3 REPLIES 3
Ayman Altounji
Valued Contributor

Re: High CPU Utilization

I have 5 Compaq ML530's running NetWare 4.11/SP9 with very similar symptoms. CPU utilization behaves strangely, and if left alone, the server will eventually stop sending and receiving on the network. Here is the progression of symptoms as we have observed them:

1. When the server is started, CPU utilization and all services function normally.

2. The first symptom is when MONITOR.NLM shows CPU utilization sticking near a percentage, with minor fluctuations. For example, one server rose to approximately 70%, only occasionally bouncing higher, but never dropping to 0% (it behaves as if 70% = 0%). There is no specific server process/NLM that is associated with the high CPU utilization -- the only correlation is "histogram overhead time."

3. Next, CIM32 loses contact with the CIM agent on the server. Yet all other services appear unaffected -- the server can be PINGed, and it still services file/print requests.

4. CPU utilization goes to 100% and/or 0% and sticks (no fluctuations). At this point, we've seen problems with some file/print requests -- such as some printers unable to attach to their queues. However, the server is still (mostly) operational. Also, we've noticed that "No Available ECB count" begins to slowly increment, and the OS starts to allocate additional Packet Receive Buffers.

5. Terminal symptom -- the server basically drops off of the network. It can no longer be PINGed, and is unable to service any file/print requests. At the server console or via a RIB board, the OS still responds to requests. You can switch screens, etc. -- it's as if the NIC drivers are unloaded. MONITOR reports bytes in/out are no longer incrementing. The server must be rebooted at this stage to restore operation.

Since our problem started, some of the hardware on these 5 servers have changed; but they all started out as ML530's with 5300 array controllers, and Compaq's 3123 ethernet NICs.

We've moved the array and NIC's to different slots, IRQ's, etc., with no positive results as yet.

The Compaq SURVEY utility and the Novell utility CONFIG.NLM seem to exacerbate the problem also. For example, once symptoms start, the CONFIG.NLM will load, but never automatically unload itself (it is supposed to write its report to a file and unload).

If anybody has any insight into these problems, please post here. Thanks!
Ayman Altounji
Valued Contributor

Re: High CPU Utilization

Load Monitor/Server Parameters/NCP/Set Client File Caching Enabled = Off. This will disable something the client is doing that locks the file. The server must be rebooted after this is set. We had the exact same problem and this cleared it up. I also heard there is a new FILESYS.NLM in FLSYSFT2.EXE on the Novell site.
Ayman Altounji
Valued Contributor

Re: High CPU Utilization

We have had a number of servers doing this. Uninstalling the Compaq Management Agents and/or Toolbox(if you are using it) stop the problem from happening.