gianm opened a new pull request, #20825: URL: https://github.com/apache/druid/pull/20825
Two main changes: First, Dart workers (Historicals) no longer look for Brokers in service discovery. Instead, they consider a Broker gone (and cancel local work) when they haven't heard from it within the timeout `druid.msq.dart.worker.controllerTimeout`. This is an improvement because it allows graceful shutdown of Brokers. (Brokers remove themselves from discovery at the beginning of the graceful shutdown window.) Second, Brokers now sync running query IDs (via DartWorkerSyncer) to Historicals every `druid.msq.dart.controller.syncPeriod`. Because this provides workers with a way to detect and remove defunct queries, there is no longer a need for infinite retries on controller-to-worker RPCs. This is also a step towards moving the Broker away from using service discovery to detect worker liveness. It will be able to use the periodic syncs to do so. -- 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]
