Github user StephanEwen commented on the issue: https://github.com/apache/flink/pull/3539 @shuai-xu Had a look at the changes. In general, I like the decoupling of failover handling from the execution graph. Some suggestions: - How about we rename the `FailoverCoordinator` to `FailoverStrategy`? - We should probably implement one very simple failover strategy (restart all) and one better one (failover region based). That has the advantage of having a default implementation that is simple and robust (same as now) for all conservative users. - I would suggest to let commands like `fail()` and `cancel()` that go directly to the `ExecutionGraph` stay on the execution graph. They are meant to apply the whole graph and not be specially handled. The failover coordinator could only handle failures from individual tasks.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---