Hello, First Open System profile:
Go to Incidents and create new incident rule:
Put name of your incident and on Conditions tab click add:
Click create measure on and on host performance find Total CPU Time. Setup there your own measure name, warning and severe thresholds. Severe thresold will be de one which will fire your incident:
(values are only example, put there what you need). Click Add => Close => Ok and you will go to Incident window with measure picked:
Than go to Actions => Advanced Configuration. You will be able to pick action plugin which will be fired with incident:
Set duration of sampling according to your needs (I don't recommend to long times on production environments). If you want to store those samplings until manual delete (dynatrace is overwriting old data after storage is full), you can check lock sesion. Click Add => Create and now this rule will work for all hosts in your environment. If you want to filter it to only one you can try different options. Please check Appmon documentation for more information about Incident rules.
Hi sebastian k.,
I’m afraid that knowing the method names isn’t enough. In our environment, many (WAR) packages use the same method/class name. And some methods, as shown in your screenshot, are Java’s foundation methods which almost every package use. Unless Dynatrace has a feature to help us on tracing back to the troubled package in this CPU sampling screen, it would be really hard for us to know where the problem is.
Thread dumps can also be helpful to analyze what is hapening with application. But generally thread dumps by me are used when I see that cpu i fine but on purepaths there is huge elapse time while starting new thread. This can tells us that there is too much waited threads or thread pool is too small. In such case thread dump is helpful. If you have cpu spike, cpu sampling is best what you can get. Please tell me more about problems you have, maybe I will be able to suggest some solution.