gaoyunhaii commented on a change in pull request #17:
URL: https://github.com/apache/flink-ml/pull/17#discussion_r731780041
##########
File path:
flink-ml-iteration/src/main/java/org/apache/flink/ml/iteration/operator/HeadOperator.java
##########
@@ -129,12 +168,89 @@ public void setup(
public void initializeState(StateInitializationContext context) throws
Exception {
super.initializeState(context);
+ parallelismState =
+ context.getOperatorStateStore()
+ .getUnionListState(
+ new ListStateDescriptor<>("parallelism",
IntSerializer.INSTANCE));
+ OperatorStateUtils.getUniqueElement(parallelismState, "parallelism")
+ .ifPresent(
+ oldParallelism ->
+ checkState(
+ oldParallelism
+ == getRuntimeContext()
+
.getNumberOfParallelSubtasks(),
+ "The head operator is recovered with
parallelism changed from "
+ + oldParallelism
+ + " to "
+ + getRuntimeContext()
+
.getNumberOfParallelSubtasks()));
+
+ // Initialize the status and the record processor.
processorContext = new ContextImpl();
- status = HeadOperatorStatus.RUNNING;
- recordProcessor = new
RegularHeadOperatorRecordProcessor(processorContext);
+ statusState =
+ context.getOperatorStateStore()
+ .getListState(new ListStateDescriptor<>("status",
Integer.class));
+ status =
+ HeadOperatorStatus.values()[
+ OperatorStateUtils.getUniqueElement(statusState,
"status").orElse(0)];
+ if (status == HeadOperatorStatus.RUNNING) {
+ recordProcessor = new
RegularHeadOperatorRecordProcessor(processorContext);
+ } else {
+ recordProcessor = new TerminatingHeadOperatorRecordProcessor();
Review comment:
If it is previously finished and then taking a checkpoint, but after
that failover happen and it restarted from the latest checkpoint.
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
