Github user tillrohrmann commented on a diff in the pull request:
https://github.com/apache/flink/pull/2436#discussion_r76645756
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/rpc/taskexecutor/TaskExecutorToResourceManagerConnection.java
---
@@ -93,22 +94,24 @@ public void start() {
registration.startRegistration();
Future<Tuple2<ResourceManagerGateway,
TaskExecutorRegistrationSuccess>> future = registration.getFuture();
+
+ ExecutionContext executionContext =
taskExecutor.getRpcService().getExecutionContext();
future.onSuccess(new OnSuccess<Tuple2<ResourceManagerGateway,
TaskExecutorRegistrationSuccess>>() {
@Override
public void onSuccess(Tuple2<ResourceManagerGateway,
TaskExecutorRegistrationSuccess> result) {
- registeredResourceManager = result.f0;
registrationId = result.f1.getRegistrationId();
+ registeredResourceManager = result.f0;
}
- }, taskExecutor.getMainThreadExecutionContext());
+ }, executionContext);
--- End diff --
If we establish the contract that you first check
`registeredResourceManager` before accessing `registrationId` it should be
thread-safe wrt the `registeredResourceManager` and `registrationId` fields. We
could do that more explicitly by returning `null` in `getRegistrationId` if
`registeredResourceManager` is not set.
Additionally, we would have to guard the `start` method which should be
doable without extra costs since it will be started only by one thread anyway.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
