high cpu utilzation due to management agents

[Ayman Altounji]

CPU utilization at 100% after boot. CPQMON shows the following process active initially: cmafcad -c /dev/casa -p 60
Later, (about 60 seconds) the rest of the management agent processes appear. I/O wait is at 25%, CPU utilization remains at 75%.
This continues until memory allocation errors are posted: crom: copyin_all_reg(): sptalloc() failed. Couldn't allocate memory. At this time server is unusable and must be rebooted. This happens every boot. The server is rendered useless.
Removing all Compaq Management Agents (version 4.9) eliminates this problem.
Is there a solution to this problem that will allow me to run the Management Agents?
EFS is 5.38b.

[Ayman Altounji]

Disabling only cmafca prevents the problem...but there is no fiber channel in use on this machine.

[Ayman Altounji]

Another little nasty side effect of this problem is there is no longer any free space on the root filesystem (all 1gb used up). It seems the Fibre Channel Agent was busy filling up my drive while I was working on the problem. I cannot locate the file (error log?) it was writing to. Any ideas?

[Ayman Altounji]

The 500 MB file (filename core) was found at /usr/local/compaq/stor/etc

[Ayman Altounji]

Please advise how to prevent this trouble so that server integration does not require the additional step of disabling the Fibre Channel Agent. Note: I have 2 ML370's and 2 DL380's. This problem exists on only one of the DL380s. The only obvious difference is that this is the only server with a 53xx Array controller (5304). The integrated raid chip is also present (removing it makes no difference). There are no other PCI cards installed. Finally, as UNIX boots, a warning is posted stating UNKNOWN PCI CONFIG DEVICE ID 00000010. Any ideas would be appreciated.

[Ayman Altounji]

I restored the system to previous working state. The PCI CONFIG SPACE problem was corrected (suspected symbios slha driver that was added with installpkg but that could not be removed with removepkg - SCO problem). I boot the system and cpu utilization was 0%. I tried to stop the FCA agent and received the msg that this agent was NOT running. Then I rebooted to single-user mode, installed efs 5.38b, rebooted, and now I get 100% cpu utilization because of cmafca. Going back to EFS 5.38a does NOT correct the problem. I have to remove S99cmafca from /etc/rc2.d.
Seems that EFS 5.38b is broke.

[Ayman Altounji]

Install the 4.90 Management agents that are separate from EFS 5.38b. You can download the latest Management agents from this web address.



http://www.compaq.com/support/files/server/us/locate/26_1121.html



Only install the managent agent software for the hardware that is actually on your system. Otherwise the management error logs will fill up and hog the CPU, trying to monitor hardware that does not exist.