Github user zentol commented on the issue: https://github.com/apache/flink/pull/5718 I don't want to make a fuss over a single test, so for the purposes of this test it's fine to just remote the timeout. In general however i think it would be pretty neat if we would add such a `Timeout` to the `TestLogger` class. I see the problem of debugging tests that have timeouts, but removing timeouts altogether is the wrong conclusion imo. Instead, we could implement the `Timeout` such that it doesn't fail the test if a certain profile/property is set. With this we keep the benefits of test timeouts (CI service independence, less special behavior locally vs CI, fixed upper bound for test times which is particularly useful for new tests, "guarantee" that the test terminates) while still allowing debugging. In fact we may end up improving the debugging situation by consolidating how timeouts are implemented instead of each test rolling their own solution that you can't disable. The travis watchdog _cannot_ be removed as it covers the entire maven process that from time to time locks up outside of tests. That said, the fact that we have to rely on the travis watchdog _to ensure that tests terminate_ is a bad sign. Not to mention that it already forced us to introduce workarounds to make tests "compatible", like the kafka tests and others that print stuff for the sole purposes of not triggering it.
---