Dynatrace High CPU consumption

[SachinJindal]

Hello

we have Dynatrace installed on Siebel Webservers(Linux) and we have seen around 20% of the CPU consumption increment after the Dynatrace enabled for RUM, has anyone has seen this issue ? and anything can be for optimization?

Regards

Sachin

[benjamin_mullan]

Not related to Siebel in particular but the only times I have seen high CPU utilisation in my linux servers is when Trend/McAfee security software causes really high cpu util.

Could be completely unrelated to your situation but worth knowing that those sort of inspection type software can introduce issues. If you have a test box try turning these off if you have them.

Which process is increasing? Dynatrace it self or one of the other processes? 

[SachinJindal]

Apache Webserver CPU consumption is going high, lets say without DT its consuming 30% on peak load but with DT it is reaching more than 50%

Thanks

[benjamin_mullan]

@SachinJindal I found a thread from a while ago seems like 3-5% is acceptable, and only the network module is able to throttle.
https://community.dynatrace.com/t5/Dynatrace-Open-Q-A/Dynatrace-Agents-Overhead/m-p/51071

I would open a support ticket with a diagnostic dump. https://dt-url.net/lb626l9 unless someone comes back with issues explicitly with Apache.

Out of curiosity what sort of request throughput are you having on this service?  and whats the version and Process technology name? 

 

[Avinash1]

@benjamin_mullan 

Hi,

I could see CPU is getting high on AG server, one of the service is consuming more CPU on the server.

What could be Exact reason behind this CPU high utilisation and where should I look for the RCA for this issue 

On console I can see only CPU usage but not able to find out the reason behind it ?

Plz suggest

[Julius_Loman]

@SachinJindal If there is not a high CPU related to oneagents processes (oneagentos, extensions, etc..) I recommend disabling RUM on the webserver process groups to check if the CPU drops. If it's RUM then is more about diagnosing the issue there - could be related to some specific setting such as injection rules (if you have them). Do you have also log enrichment enabled? Definitely 20% increase seems too high. What is the traffic?

[SachinJindal]

Hello @Julius_Loman  if we disable RUM than more likely we are only doing infra monitoring which works perfectly as well. also tried couple of apache setting disabled in Dynatrace monitoring time queueing etc , as soon as i will disable httpd deep monitoring it comes back to normal CPU.

[Julius_Loman]

Now it's about finding your why it's such overhead. Keep deep monitoring and disable just RUM sensor by turning off RUM on the process group level.

[SachinJindal]

on process group i can only see below option.

[Julius_Loman]

The following toggle:

[SachinJindal]

Thanks , This test case we already tried and disabling RUM it comes back to normal CPU.

[Julius_Loman · Accepted solution]

@SachinJindal  OK, then this is caused by the RUM sensor. I'd try defining injection rules to put the Dynatrace RUM only on the subset of pages and see how much the CPU usage increases. Without knowing your load it's difficult to say what is causing the issue. Maybe RUM sensor is injecting also into responses where it's not necessary or desired.  Also in parallel to get this resolved - open a support ticket or use the built-in product chat for assistance.

[SachinJindal]

Thanks for the suggestion and provided the perf tool stats to support team for analysis.