I need your help again.
With our agent enable our transactions are 41% more slow.... Its too much
I already modified IntroscopeAgent.profile and disabled stuff like: Socket, UDP, Network tracing, sqlagent tracing and deep inheritance.
On pbl files and pbd files what can we disable or configure to speed up at least 10% more fast?
The process to identify where the Agent overhead is coming from and reduce it is usually fairly easy.
What you need to do is:
1) Go to the Investigator Tree
2) Select your Agent
3) Select the “Search” Tab
4) Type “Responses Per Interval”
5) Sort by largest values first
The largest ones are where the overhead is coming from, they mean we’re running tracers on methods that have very high invocation rates and usually very small response times, and are therefore not really interesting to measure, so these are the ones you will want to turn off.
Have you been able to create a Glassfish package in APM Command Center by the way?
We are studying if the measures can be turned off.
Yes i created a Glassfish package in APM Command Center.
Thanks for your support.
I'll replay more later if i can turn off or no those measures.
Yes, please respond on the other post about Glassfish. Given that these types of issues can get very involved, opening a case may be the quickest way to go. But for the moment, let's continue to work the issue through the Community.
Based on Florian's suggestion, were you able to turn off additional tracers that may be impacting the agent? Looking at your IntroscopeAgent.profile would be helpful to determine next steps. Please post it if Florian's suggestion does not resolve/improve this issue.
Also what is your directive line in the agent profile? Any custom pbd's or pbl's? Anything in hotdeploy?
We dont have any custom pbd´s or pbl´s. Nothing in hotdeploy.
Hi Fernando: It sounds like you have a clear next step. So marking as answered. However, after your discussion with the customer and following Florian's and Matt's suggestion, if the problem persists , please post what was done and current status. Thanks
What version of APM are you running?
Which agent (Java, .NET)?
If it is the Java agent, which version and vendor is the JVM?
In version 10.0 we hit an issue with the JVM version 1.6 versus 1.7. With 1.6, we had to use the IntroscopeAgent_noredef.profile so that the instrumentation wouldn't try to re-instrument the classes.
When we had the 1.7 configuration on a 1.6 JVM, the CPU was rather high, more than 10% when it was typically under 5%.
Hope this helps,
Thanks for your info. I will analyse here.
On production we have some servers with agent version 10.0, i will check the JVM versions on that machines.