wpc009 commented on a change in pull request #18475:
URL: https://github.com/apache/flink/pull/18475#discussion_r794346464
##########
File path:
flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/io/StreamMultipleInputProcessor.java
##########
@@ -157,4 +186,66 @@ private void fullCheckAndSetAvailable() {
}
}
}
+
+ /** Visible for testing only. Do not use out side
StreamMultipleInputProcessor. */
+ @VisibleForTesting
+ public static class MultipleInputAvailabilityHelper {
+ private final CompletableFuture<?>[] cachedAvailableFutures;
+ private final Consumer[] onCompletion;
+ private CompletableFuture<?> availableFuture;
+
+ public CompletableFuture<?> getAvailableFuture() {
+ return availableFuture;
+ }
+
+ public static MultipleInputAvailabilityHelper newInstance(int
inputSize) {
+ MultipleInputAvailabilityHelper obj = new
MultipleInputAvailabilityHelper(inputSize);
+ return obj;
+ }
+
+ private MultipleInputAvailabilityHelper(int inputSize) {
+ this.cachedAvailableFutures = new CompletableFuture[inputSize];
+ this.onCompletion = new Consumer[inputSize];
+ }
+
+ @VisibleForTesting
+ public void init() {
+ for (int i = 0; i < cachedAvailableFutures.length; i++) {
+ final int inputIdx = i;
+ onCompletion[i] = (Void) -> notifyCompletion(inputIdx);
+ }
+ }
+
+ public boolean isInvalid() {
+ return availableFuture == null || availableFuture.isDone();
+ }
+
+ public void resetToUnavailable() {
+ availableFuture = new CompletableFuture<>();
+ }
+
+ private void notifyCompletion(int idx) {
+ if (availableFuture != null && !availableFuture.isDone()) {
+ availableFuture.complete(null);
+ }
+ cachedAvailableFutures[idx] = AVAILABLE;
+ }
Review comment:
`notifyCompletion` may competing with `checkReusableAndReset`. We could
not prevent that without using synchronisim locks.
But, this competing may be not harmful.
In your fix, the `AtomicReference` can not prevent that either. The
`isDone()` check is still seperate with the reset.
Both ours fix is working. I have test them with 300 test runs. Did not see
anykind of hang-up.
--
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]
