Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/5427#discussion_r184477785
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointCoordinator.java
---
@@ -1009,6 +1013,11 @@ public boolean restoreLatestCheckpointedState(
LOG.debug("Status of the shared state registry after
restore: {}.", sharedStateRegistry);
+ // Instruct the master hooks to initialize their state
(unconditionally)
+ LOG.debug("Initializing the master hooks.");
--- End diff --
Can you elaborate a bit why this initialization is happening in all cases?
An alternative would be to have a `reset()` method or so on the master hook
that is called further below, in the `if (latest == null)` code block.
Initializing the state seems a tad bit unituitive to me here - I somehow
assume that an init function is called once, while this one here is called on
every restore.
---
