IBM Tivoli Monitoring, Version 6.3

CPU throttling troubleshooting

Problems can occur with the CPU throttling of the monitoring agent.

Table 1. CPU throttling problems and solutions
Problem	Solution
While using CPU throttling, a review of the log file does not indicate if the CPU usage limits on the monitoring agent are limiting events.	The Log File Status attribute group shows information about each log file that is being monitored. In particular, for each file that is being monitored, the difference between the Current® File Position and the Current File Size show the amount of data the log file monitor has yet to process. This amount of data that remains to be processed, gives an indication of how far behind the log file agent is. Setting the `KBB_RAS1=ERROR (UNIT:kumpcpu STATE)` will create a trace entry every time a thread is actively slowed. The trace entry shows the number of process seconds used, the elapse time, and the utilization. The wait time in milliseconds is the number of milliseconds needed to delay meeting the CPU throttling goal. The number of hot threads are the number of active threads currently doing CPU-intensive work. Example output: `Spent 0.589910000 sec processor, 5.890441194 sec elapsed, 5.0073%; Waiting 8 millisec; 1 hot threads; From (kumpiafi.c,1161"KUMP_ConstructAttrFilterInfo")`
While using CPU throttling, you cannot see the CPU usage for the agent.	Setting the `KBB_RAS1=ERROR (UNIT:kumpcpu METRICS STATE)` creates a trace entry every minute showing the processor utilization of each process. All platforms can determine the processor utilization of a process. For example, the Windows Task Manager or the Linux top command can be used to determine the process information. Example output: `Process processor utilization 3.6311% (8.71 / 120.00); Process interval utilization 3.6311% (8.71 / 120.00); 2 cores`
While using CPU throttling, you cannot see if the CPU usage limits for the agent are exceeded.	An error trace is issued when the CPU throttling maximum is exceeded for an interval. If the processor utilization is within 10% of the goal, it is considered an informational trace message. If the processor utilization exceeds the goal by 10%, it is considered an error trace message. Example output: `Informational: Interval processor utilization 11.1372% (26.72 / 120.00) exceeds processor utilization goal 10 with 2 cores`

Feedback