Github user tillrohrmann commented on a diff in the pull request:
https://github.com/apache/flink/pull/1153#discussion_r41127760
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/leaderelection/LeaderElectionService.java
---
@@ -67,4 +67,20 @@
* @return true if the associated {@link LeaderContender} is the
leader, otherwise false
*/
boolean hasLeadership();
+
+ /**
+ * [EXPERIMENTAL] Returns <code>true</code> if the {@link
LeaderContender} is leader. A call
+ * to this method might block.
+ *
+ * <p>This forces a synchronous check at the respective state backend.
It is possible
+ * that is does not reflect the current state at the {@link
LeaderContender}, which is notified
+ * asynchronously. Therefore it is possible that {@link
#hasLeadership()} and {@link
+ * #syncHasLeadership()} have different return values.
+ *
+ * @TODO @tillrohrmann Is it OK to collapse this with {@link
#hasLeadership()}?
+ *
+ * @return true if the associated {@link LeaderContender} is the
leader, otherwise false
+ */
+ boolean syncHasLeadership();
--- End diff --
I do not know, whether this method adds much value. The problem is that
reading the leader path data and comparing the leader session ID do not happen
atomically. Thus, the moment you've just read the leader path, it can happen
that the node's data and, thus, the leader changed. IMO, this is the same
problem as with `leaderLatch.hasLeadership`.
---
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.
---
