Diagnosing high CPU usage on a distributed system

You can diagnose that CPU usage is high on a distributed system by determining whether a monitoring component, application, or process running on the system might be the cause of the problem. Also, you can use the ITMSUPER tools, such as the Connectivity tool to provide more information.

Preliminary diagnostics

1. Determine whether an IBM® Tivoli® Monitoring component is the root cause. Another application or process running on the system might be causing high CPU usage.
2. Use the tools and data provided by Task Manager to identify the process causing high CPU usage. In the Processes tab you can reorder the processes by CPU usage. An example of a process name is kntcma.exe for the Windows OS agent.
3. Use the top command to display processes using high CPU. For UNIX, you can also use the ps auxww command.
4. Verify the following:
   • Is historical data collection enabled?
   • Is the database undergoing a backup?
5. Is the situation writing a lot of event logs?
   • Yes: Disable all event log monitoring situations.
6. Select each of the workspaces in turn, to see which one is consuming high CPU.
7. Running the following ITMSUPER tools might also provide more information:
   • Stressed Resources tool
   • Connectivity tool
   • Situations tool
8. When the computer (where the monitoring agent is running) has multiple Network Interface Cards (NICs), the agent might not be bound to the Primary NIC. The agent might not be able to establish connectivity with the monitoring server. High CPU usage can result from the agent's frequent attempts to connect.
   1. To correct this, you might need to set the environment variable KDEB_INTERFACELIST = '!*' or KDEB_INTERFACELIST = IP_address, where IP_address is the address of the NIC.
   2. Make the changes in the associated agent *ENV configuration file for Windows, or the *.ini configuration file for UNIX or Linux.