jrudolph commented on issue #422: URL: https://github.com/apache/incubator-pekko-http/issues/422#issuecomment-1903924454
I don't know about any change which could have caused this issue. Can you still validate that the old version based on Akka is still fine (to rule out a coincidence where something else in the wider environment might have changed in the meantime)? Your best bet is trying to collect debug log message from `org.apache.pekko.http.impl.engine.client.pool.NewHostConnectionPool` to isolate the issue. Since the issue indeed looks like it could be caused by the infamous stream cancellation race condition, you could try setting `stream-cancellation-delay` to much higher values to rule out any race-conditions that might still occur in a 1 second time frame caused by GC or other interruptions. What kind of requests run into that problem? Usually, idempotent requests are retried a few times, which would make it even more unlikely to happen. So, this might also be an issue of unfortunate timing, where an non-idempotent request (e.g. POST) is run at exactly the same time as the server decides to close the connection (however, that scenario is reported in an unclear way caused by stream cancellation). In that case, playing around with `idle-timeout` and `host-connection-pool.keep-alive-timeout` (see also its documentation in the `reference.conf`) might help mitigate the issue. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
