We are running YourKit 2020.9 and remote profiling a Java application hosted inside a Linux Docker container. Checking /proc/stat and top shows that our CPU has idle time, and is usually sitting at around 60-70% usage.
We used the guide https://www.yourkit.com/docs/java/help/docker.jsp to setup our Dockerfile and have no special options enabled, everything default.
When attached to the remote agent, it shows constant 100% CPU usage. System is at 100% and Process + Kernel is at 100% always for the entire duration of the session.
This is definitely incorrect, is there any way to fix this or improve the accuracy of the estimate?
Thanks.
System and Process + Kernel CPU usage estimates stuck at 100
-
- Posts: 1
- Joined: Tue Dec 01, 2020 5:52 pm
-
- Posts: 6172
- Joined: Wed Aug 11, 2004 8:37 am
Re: System and Process + Kernel CPU usage estimates stuck at
Hello,
It's a new issue.
What processor count is shown in the "Summary" tab, in the "Operating System" section?
Does it match the actual number?
Could you please send to [email protected] the following:
- the profiler agent log file <user home>/.yjp/log/<session name>-<pid>.log
- a performance snapshot file demonstrating the problem
- "uname -a" output of the host
- "uname -a" inside the container.
It's a new issue.
What processor count is shown in the "Summary" tab, in the "Operating System" section?
Does it match the actual number?
Could you please send to [email protected] the following:
- the profiler agent log file <user home>/.yjp/log/<session name>-<pid>.log
- a performance snapshot file demonstrating the problem
- "uname -a" output of the host
- "uname -a" inside the container.
-
- Posts: 6172
- Joined: Wed Aug 11, 2004 8:37 am
Re: System and Process + Kernel CPU usage estimates stuck at
Update:
The problem was caused by using particular OpenJDK version which incorrectly reported the processor count.
The problem was solved by using another OpenJDK version (namely, from https://hub.docker.com/r/adoptopenjdk/openjdk8).
The problem was caused by using particular OpenJDK version which incorrectly reported the processor count.
The problem was solved by using another OpenJDK version (namely, from https://hub.docker.com/r/adoptopenjdk/openjdk8).