Documentation feedback
In a future variant of the troubleshooting section, it could elaborate about hands-on details, like how to investigate CrateDB's throttling mechanisms.
One could enable logging to see when the related throttle mechanism kicks in by:
SET GLOBAL 'logger.io.crate.execution.engine.indexing' = 'DEBUG';
Or, if this creates too much noise, more concrete:
SET GLOBAL 'logger.io.crate.execution.engine.indexing.BatchIteratorBackpressureExecutor' = 'DEBUG';
Afterwards there should be log entries like:
Pausing consumption jobId={} delayInMs={}
Resuming consumption jobId={}
This may help to understand if throttling is valid or not.
Please note that mostly the throttling is not just kicking in because of CPU load, but also, or even more likely, because of memory/heap saturation.
Thanks, @seut.
Documentation feedback
In a future variant of the troubleshooting section, it could elaborate about hands-on details, like how to investigate CrateDB's throttling mechanisms.
Thanks, @seut.