lhotari commented on issue #24695: URL: https://github.com/apache/pulsar/issues/24695#issuecomment-3249151897
Now I see the correlation. When `maxPendingPublishRequestsPerConnection` value is reduced, the consuming will be more responsive with the tradeoff that latency increases and throughput goes down for both producing and consuming. In the above case, I had increased `maxPendingPublishRequestsPerConnection` to `10000` from the default `1000` to improve producing throughput. This seems to be a problem where the produced entries and the consuming (dispatching) are constrained on the same thread and get queued up in the same queue. There's also #23865. Currently `readMoreEntries` gets called in random Pulsar IO threads, which doesn't make much sense. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
