@ganncamp Thank you. In the meantime I may have found the root cause. I’m saying may because the whole-project analysis now worked 4 times, and considering that it was only working 1/7 times before this seems quite unlikely to not have had an effect, but I’m not sure if it fixed all the issues.
A new failure popped up which lead me to this thread: Fail to execute es requestES refresh request on indices 'components' - #8 by shiva_sq
I tried upping the elasticsearch memory to 4GB, by using sonar.search.javaOpts=-Xmx4G -Xms4G -XX:+HeapDumpOnOutOfMemoryError.
Since I was now getting OOM on startup I scaled down the heap size for the compute engine and the web engine. They were both set to 32GB. Now I set:
jvmOpts: "-Xmx16G -Xms1G -XX:+HeapDumpOnOutOfMemoryError"
jvmCeOpts: "-Xmx16G -Xms1G -XX:+HeapDumpOnOutOfMemoryError"
Hint: If you’re reading this and not using Helm, I think these should be sonar.web.javaOpts and sonar.ce.javaOpts. In the Helm file there’s currently no setting for sonar.search.javaOpts so I had to add that to the normal settings.