lukecwik commented on a change in pull request #11591:
URL: https://github.com/apache/beam/pull/11591#discussion_r418756561
##########
File path:
runners/flink/src/main/java/org/apache/beam/runners/flink/translation/functions/FlinkExecutableStageFunction.java
##########
@@ -247,25 +247,27 @@ public void reduce(Iterable<WindowedValue<InputT>>
iterable, Collector<RawUnionV
timerInternals.advanceSynchronizedProcessingTime(BoundedWindow.TIMESTAMP_MAX_VALUE);
// Now we fire the timers and process elements generated by timers (which
may be timers itself)
- try (RemoteBundle bundle =
- stageBundleFactory.getBundle(
- receiverFactory, timerReceiverFactory, stateRequestHandler,
progressHandler)) {
-
- PipelineTranslatorUtils.fireEligibleTimers(
- timerInternals,
- (KV<String, String> transformAndTimerId, Timer<?> timerValue) -> {
- FnDataReceiver<Timer> fnTimerReceiver =
- bundle.getTimerReceivers().get(transformAndTimerId);
- Preconditions.checkNotNull(
- fnTimerReceiver, "No FnDataReceiver found for %s",
transformAndTimerId);
- try {
- fnTimerReceiver.accept(timerValue);
- } catch (Exception e) {
- throw new RuntimeException(
- String.format(Locale.ENGLISH, "Failed to process timer: %s",
timerValue));
- }
- },
- currentTimerKey);
+ while (timerInternals.hasPendingTimers()) {
Review comment:
Doesn't it make sense to make this change in batch as well for
spark/flink in FlinkExecutableStageFunction and SparkExecutableStageFunction?
Any watermark based timers should continue to be eligible and continue to
fire while in batch while the processing time timers should be dropped.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
