cpu usage
Answered
I know there are some posts from a long time ago about this but I'm using the latest version via docker images and I was a little surprised to see this while evaluating teamcity. I need a simple(ish) CI tool for a small team and when I booted this up, despite not having any projects or jobs to do the tool takes over the entire cpu for a good 5 minutes roughly every time it first runs. It does calm down later but I was wondering if there is some configurations I could do to make it less of a cpu hog?
Please sign in to leave a comment.
Hi, this does not look like normal behavior. You have four cores, which should be enough. The memory setting (-Xmx768m) may be a bit low for a 64-bit JVM, but let's not get ahead of the investigation.
Could you please take ten server thread dumps with approximately a 5-second interval when the issue takes place? If the UI is not responsive, you can take thread dumps manually with the help of jstack. Additionally, we will need a CPU profiling snapshot; we have a short article that describes how to get it. Once all data is collected, please upload it along with server logs to https://uploads.jetbrains.com/.
It seems to be limited to just the first 5-10 minutes after I boot up. Which was often while I was configuring the docker images etc but once I got all that sorted and left it running for a bit it smoothed right down. A bit low indeed but it was originally a much smaller vm just to run the test. I bumped it up to get the 4 cores so more ram is possible now but I'll leave it at this for now. I also am using the internal database currently just for ease. Should I still collect the data because I noticed on the cpu profiling page you linked it mentioned to wait 20 minutes before doing it. So I assume a heavy load is expected at the beginning?
Hey Robert,
I discussed this internally, and yeah, high CPU usage is expected on startup because the server compiles DSL and JSP. Regarding the xmx setting, the minimum value for 64-bit Java is -Xmx1024m. The recommended setting for a medium server is -Xmx2048m. I suggest you increasing the value at least to 1024m. Otherwise, the JVM may have unnecessary interruptions for the garbage cleaning process.