Github user tillrohrmann commented on a diff in the pull request:
https://github.com/apache/flink/pull/5730#discussion_r177778690
--- Diff:
flink-tests/src/test/java/org/apache/flink/test/example/client/JobRetrievalITCase.java
---
@@ -80,42 +95,38 @@ public void testJobRetrieval() throws Exception {
final JobGraph jobGraph = new JobGraph(jobID, "testjob",
imalock);
- final ClusterClient<StandaloneClusterId> client = new
StandaloneClusterClient(cluster.configuration(),
cluster.highAvailabilityServices(), true);
-
// acquire the lock to make sure that the job cannot complete
until the job client
// has been attached in resumingThread
lock.acquire();
--- End diff --
In Flip-6 there is no such concept as attaching to a running job anymore.
What attaching effectively did in the legacy code, was to tell the
`JobManager` that there is again a client waiting to be notified about the job
result.
In Flip-6 the job result will be stored on the `Dispatcher` also if there
is no client waiting to be notified. This allows the client to request it at
any time (modulo some cache size cleaning conditions).
---
