We installed OneAgent on the customer's AIX 7.2 server and they reported to us the server's TPS and response time were affected (Server TP and response time lowered by about 20% after installing OneAgent).
We show the CPU and RAM, that OneAgent consumed during testing time. It showed normal values (not too high)
Has anyone faced with this issue? Or any advice for debugging this issue?
Thanks for your support!
I understand that during the tests OneAgent did not affect the server performance? Please write whether the load is constant or persists for a few minutes after startup? (This situation can occur as the first instrumentation is performed by OA).
As a first step, switch OA to infra-only mode and let us know if the load has decreased:
Thanks for your support. I tested following your suggestion with both modes: Fullstack & Infrastructure. Below about the test results:
- How we test: We generated requests from the testing machine to Monitored host server (AIX) in 15 minutes.
- About CPU/RAM/Memory/Network: In both cases, there was no big difference.
- About Server TPS and Response time: Fullstack mode increased by 18%-20% response time, compared with Infrastructure mode.
Do you have any advice for optimizing OneAgent in full-stack mode to fix the response time issue?
Thanks in advance,
When Full-Stack mode causes an overhead problem, there can be many causes. Probably ie it is some interference on the side of source code instrumentation. It could be some specific setting of your application or version.
I have two more questions:
In general, at the same time I recommend creating a ticket in support, as you may need to exclude specific items on the DT side or your application.
Here's also some information on how OA can affect the overhead: https://community.dynatrace.com/t5/Open-Q-A/How-much-overhead-does-the-oneagent-take/m-p/27422