Flaky unit tests about spark-jobserver HOT 3 CLOSED

@valan4ik You were right, there is indeed a race condition in the callback job test. Should be stable now.

from spark-jobserver.

@valan4ik There still seem to be some problems in JobManagerActorSpec. In Spark 3, only a single context can be running per JVM, so the error message looks a bit different, but the underlying problem should be the same. See Travis CI. I will add some similar logic to JobMangerActorSpec's before and after methods in the Spark3 PR.

But the test for Spark 2 seem to be stable now.

from spark-jobserver.

[info] - should return valid managerActorRef if context exists *** FAILED *** (1 second, 923 milliseconds)
[info]   false was not true (AkkaClusterSupervisorActorSpec.scala:345)

from spark-jobserver.