crazyzhou commented on pull request #14890: URL: https://github.com/apache/flink/pull/14890#issuecomment-783110723
Hi Stephan, sorry for the late reply for our holiday. > * Do Pravega setups always configure "ALL" as the failover strategy? We use the default `region`, it's theoretically fine because we solve this as the next answer. We actually did not care much about this before because almost all of the current Pravega Flink tests and real use cases are streaming based, so all the tasks are in the same pipelined region, so it's always a "global" recovery during a failover. > * Do the readers in the not-restarted regions throw an exception when reset to the checkpoint so that these regions implicitly also fail over? Yes, for non-restarted readers, it will throw `ReinitializationRequiredException` when reading events (ref: https://github.com/pravega/pravega/blob/f6971bd0a1213263ed9e89c644f7619ceb62603c/client/src/main/java/io/pravega/client/stream/ReaderGroup.java#L88), and then these regions will implicitly fail over. ---------------------------------------------------------------- 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: us...@infra.apache.org
