gaoyunhaii commented on a change in pull request #25:
URL: https://github.com/apache/flink-ml/pull/25#discussion_r746358395
##########
File path:
flink-ml-iteration/src/test/java/org/apache/flink/iteration/operator/ReplayOperatorTest.java
##########
@@ -50,12 +69,206 @@ public void testReplaying() throws Exception {
final int numRecords = 10;
OperatorID operatorId = new OperatorID();
+ createHarnessAndRun(
+ operatorId,
+ null,
+ harness -> {
+ // First round
+ for (int i = 0; i < numRecords; ++i) {
+ harness.processElement(
+ new
StreamRecord<>(IterationRecord.newRecord(i, 0)), 0, 0);
+ }
+ harness.endInput(0, true);
+ harness.processElement(
+ new
StreamRecord<>(IterationRecord.newEpochWatermark(0, "sender1")),
+ 1,
+ 0);
+ assertOutputAllRecordsAndEpochWatermark(
+ harness.getOutput(), numRecords, operatorId, 0);
+ harness.getOutput().clear();
+
+ // The round 1
+ harness.processElement(
+ new
StreamRecord<>(IterationRecord.newEpochWatermark(1, "sender1")),
+ 1,
+ 0);
+ // The output would be done asynchronously inside the
ReplayerOperator.
+ while (harness.getOutput().size() < numRecords + 1) {
+ Thread.sleep(500);
+ }
+ assertOutputAllRecordsAndEpochWatermark(
+ harness.getOutput(), numRecords, operatorId, 1);
+ harness.getOutput().clear();
+
+ // The round 2
+ harness.processElement(
+ new
StreamRecord<>(IterationRecord.newEpochWatermark(2, "sender1")),
+ 1,
+ 0);
+ // The output would be done asynchronously inside the
ReplayerOperator.
+ while (harness.getOutput().size() < numRecords + 1) {
+ Thread.sleep(500);
+ }
+ assertOutputAllRecordsAndEpochWatermark(
+ harness.getOutput(), numRecords, operatorId, 2);
+ return null;
+ });
+ }
+
+ @Test
+ public void testSnapshotAndRestoreOnFirstEpoch() throws Exception {
+ final int numRecords = 10;
+ OperatorID operatorId = new OperatorID();
+
+ List<Object> firstRoundOutput = new ArrayList<>();
+ List<Object> secondRoundOutput = new ArrayList<>();
+
+ TaskStateSnapshot snapshot =
+ createHarnessAndRun(
+ operatorId,
+ null,
+ harness -> {
+
harness.getTaskStateManager().getWaitForReportLatch().reset();
+
+ for (int i = 0; i < numRecords / 2; ++i) {
+ harness.processElement(
+ new
StreamRecord<>(IterationRecord.newRecord(i, 0)), 0, 0);
+ }
+
+ harness.getStreamTask()
+ .triggerCheckpointAsync(
+ new CheckpointMetaData(2, 1000),
+ CheckpointOptions.alignedNoTimeout(
+ CheckpointType.CHECKPOINT,
+
CheckpointStorageLocationReference
+ .getDefault()));
+ harness.processAll();
+
+ firstRoundOutput.addAll(harness.getOutput());
+
+
harness.getTaskStateManager().getWaitForReportLatch().await();
+ return harness.getTaskStateManager()
+ .getLastJobManagerTaskStateSnapshot();
+ });
+
+ createHarnessAndRun(
+ operatorId,
+ snapshot,
+ harness -> {
+ for (int i = numRecords / 2; i < numRecords; ++i) {
+ harness.processElement(
+ new
StreamRecord<>(IterationRecord.newRecord(i, 0)), 0, 0);
+ }
+ harness.endInput(0, true);
+ harness.processElement(
+ new
StreamRecord<>(IterationRecord.newEpochWatermark(0, "send-0")),
+ 1,
+ 0);
+ harness.processAll();
+ firstRoundOutput.addAll(harness.getOutput());
+
+ // The second round
+ harness.getOutput().clear();
+ harness.processElement(
+ new
StreamRecord<>(IterationRecord.newEpochWatermark(1, "send-0")),
+ 1,
+ 0);
+ secondRoundOutput.addAll(harness.getOutput());
+
+ return null;
+ });
+
+ assertOutputAllRecordsAndEpochWatermark(firstRoundOutput, numRecords,
operatorId, 0);
+ assertOutputAllRecordsAndEpochWatermark(secondRoundOutput, numRecords,
operatorId, 1);
+ }
+
+ @Test
+ public void testSnapshotAndRestoreOnSecondEpoch() throws Exception {
+ final int numRecords = 20;
Review comment:
It could also be 10, I changed it to 10 to be consistent.
--
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]
