Built time is slow on SonarQube

Hi, We are facing issue concerning Our SonarQube.
We recently saw that memory was a bit short on the server. After un upgrade and many controls, nothing seems to change. We are using SonarQube on the 8.6 version (build 39681).
We have check every possibility on our side (Server, Hardware, Network), and we did not find any bottleneck. Is there any known issue with this build ?
Thanks in advance for your help.
Kévin

2 Likes

Hello @klefevre
and welcome to the SonarSource Community!

Would you be able to be more explicit about what is slow with your SonarQube instance? Are the Compute Engine Tasks taking more time (long background tasks), or do you have some problems with the analysis times (sonar-scanner analysis)? What are the execution times that you are experiencing, for what projects sizes (lines of code)?
Also did you start to have this behavior after an upgrade, from which version?
Generally speaking, feel free to share more details about your platform setup and dimensioning.

And two additional questions:

  • Do you monitor your SonarQube JVMs (i.e. for heap size memory size and GC collection times)?
  • Did you apply the recommendations provided with the SonarQube Memory Tuning Guidelines?

Best regards
Sylvain

Hello @Sylvain_Combe,
Thanks for your reply.
Here are the answers to your questions:

  • Actually this is the Anaysis time that is the issue
  • We passed 10 millions lines of analysis recently,
  • No upgrade have been done on the last few months, but one is planned for August.
  • Server side, here is the configuration 8cpus, 20Go of ram, Gigabit Network, SAS 10k Disk Host,
    • There is no bottleneck on this vrtual server, everything have been check several times.
  • We did not monitor SonarQube JVMs, but we monitor all “classic” server usage (cpu, memory, disk)
  • Concerning Java Configuration, everything seems to match with recommandations. All JVMs represent approximately, 14.8G of ram usage on the server.

Thanks for your help, have a nice day !
Best regards
Kévin

Hello @klefevre
Analysis is run from your pipeline execution agents (by the scanner), a better monitoring of the SonarQube server (OS level monitoring is clearly insufficient) will not help there.
Can you share more information about one project that triggered a too long analysis time: language, size? And about your pipeline execution agent dimensioning?

When you set an analysis in verbose mode, the scanner logs allow you to the details about each rule execution times, don’t hesitate to share these findings as well.

Sylvain