reswqa commented on code in PR #21137:
URL: https://github.com/apache/flink/pull/21137#discussion_r1007555509
##########
flink-runtime/src/test/java/org/apache/flink/runtime/jobmaster/JobMasterServiceLeadershipRunnerTest.java:
##########
@@ -683,21 +665,66 @@ public void testJobAlreadyDone() throws Exception {
jobManagerRunner.getResultFuture();
JobManagerRunnerResult result = resultFuture.get();
- assertEquals(
- JobStatus.FAILED,
-
result.getExecutionGraphInfo().getArchivedExecutionGraph().getState());
+
assertThat(result.getExecutionGraphInfo().getArchivedExecutionGraph().getState())
+ .isEqualTo(JobStatus.FAILED);
}
}
+ @Test
+ void
testJobMasterServiceLeadershipRunnerCloseWhenElectionServiceGrantLeaderShip()
+ throws Exception {
+ final TestingLeaderElectionDriver.TestingLeaderElectionDriverFactory
+ testingLeaderElectionDriverFactory =
+ new
TestingLeaderElectionDriver.TestingLeaderElectionDriverFactory();
+ final LeaderElectionService defaultLeaderElectionService =
+ new
DefaultLeaderElectionService(testingLeaderElectionDriverFactory);
+
+ final JobMasterServiceLeadershipRunner jobManagerRunner =
+ newJobMasterServiceLeadershipRunnerBuilder()
+ .setJobMasterServiceProcessFactory(
+
TestingJobMasterServiceProcessFactory.newBuilder().build())
+ .setLeaderElectionService(defaultLeaderElectionService)
+ .build();
+
+ jobManagerRunner.start();
+ final TestingLeaderElectionDriver currentLeaderDriver =
+ Preconditions.checkNotNull(
+
testingLeaderElectionDriverFactory.getCurrentLeaderDriver());
+
+ final CheckedThread contenderCloseThread =
+ new CheckedThread() {
+ @Override
+ public void go() {
+ try {
+ jobManagerRunner.close();
+ } catch (Exception e) {
+ throw new RuntimeException(e);
+ }
+ }
+ };
+ contenderCloseThread.start();
+
+ // grant leadership.
+ currentLeaderDriver.isLeader();
Review Comment:
@XComp Thank you for your advice.
At the beginning, I didn't think of a good way to control the completion of
completableFuture, and the probability of deadlock reproduce in my local
environment is very high. But your suggestion did find a good control injection
point, so i decided to test it in this way.
But if we want to reproduce the problem 100%, it seems that we need another
control point. Maybe `TestingLeaderElectionDriver#isLeader` is a suitable
choice. What's your opinion?
--
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: issues-unsubscr...@flink.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
