Current CPU Load Inconsistent?
Hello All,
I'm trying to figure out what exactly Current CPU Load is measuring. I have results from top looking specifically at an IHS Web Server. The numbers returned from top don't line up to anything I have from the chart.
My chart is set to 10s Resolution, with the Current CPU Load measure aggregation set to Max. I would expect to see something in the single digit percentages in dT.
Can someone explain the differences? Thanks.
Can you try uptime command?
Here is a link: http://stackoverflow.com/questions/8491353/uptime-vs-top-cpu-usage-what-should-i-believe-why-this-difference
Best Answer
Hi Cesar
apologies, I have now double checked and my comment applied to "system load" not process current CPU load, sorry I had misread your post.
The measure you are charting should indeed apply to the monitored process so here in this case your web server. I agree that there is a discrepancy. According to top, I would expect a figure around 14% (sum of the 4 httpd processes).
One thing troubles me though. Your chart shows data every minute but you should have a data point every 10 seconds (you mentioned that's your resolution). Web server agents (I assume you have indeed deployed one inside IHS) should return data every 10 seconds. Something might be wrong with the agent/measure.
Regards
Flo
Hello Flo,
Again, thanks for your response. Concerning the data resolution: I had completely forgotten about this: As I understand it, the data points are showing only at 1 minute intervals is because this charted data was older than 1 hour, so it was being pulled from the Performance Warehouse at High Resolution, as opposed to from dynaTrace Server Memory at Full resolution.
The issue still stand as to what Current CPU Load actually measures, and, directly related, what the PROPER method is for calculating CPU usage of an Instrumented process. I'd appreciate the help if anyone knows! :)
Hi
CPU load should match the load average figure visible at the top of the "top" command (0.15, 0.18, 0.18).
Your measure on the screenshot shows values around 0.2 so it is a close match.
Regards
Flo
Hello Flo,
Thanks for the response. This does bring a couple of questions:
1. The description of this metrics is: Percentage of total CPU load generated by monitored Process.
I would assume that means the instrumented process, which in this case would be the Web Server. But, from your answer, it seems it may not be the case. Can you or anyone else offer any clarification on this?
2. Is there a measure that returns the CPU usage of the instrumented Process, which again, in this case would be the Apache HTTP Server?
Thanks.
actually I remember that top was notoriously wrong on Solaris and I had to use prstart instead but I guess you are not on Solaris?
No Solaris. We're on RHEL.
Welcome the
Dynatrace Community Forums
Check out the Forum User Guide and Forum Guidelines to learn how to get started.Community Member of the Month
August 2017
Announcing Dynatrace's Community Member of the Month for August 2017, Rick B.! Click here to read more! 
