Laboltus opened a new issue #11089: URL: https://github.com/apache/druid/issues/11089
We have historical servers with fast array of NVMe drives and 48 core CPU but frequenty I see that the bottleneck is ksoftirqd core process. As I understand it, the process is responsible for loading of memory mapped segment files into RAM and there is limitation 1x ksoftirqd per NUMA node. In our case I we have 2 NUMA nodes and I see 2x ksoftirqd processes which consume 2x 100% of cpu core at the beginning of processing of large request. All other cores are iddle druing the phase. Is there something I can do to avoid this limitation ? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
