Github user srdo commented on a diff in the pull request:
https://github.com/apache/storm/pull/1924#discussion_r99642608
--- Diff:
external/storm-kafka-client/src/main/java/org/apache/storm/kafka/spout/KafkaSpout.java
---
@@ -240,7 +240,9 @@ private boolean commit() {
private boolean poll() {
final int maxUncommittedOffsets =
kafkaSpoutConfig.getMaxUncommittedOffsets();
- final boolean poll = !waitingToEmit() && numUncommittedOffsets <
maxUncommittedOffsets;
+ final boolean poll = !waitingToEmit() &&
+ (numUncommittedOffsets < maxUncommittedOffsets ||
+ !retryService.retriableTopicPartitions().isEmpty());
--- End diff --
No, I don't think so. The tuples need to be both failed and ready for retry
for this to be true, and if it's true, the consumer will seek back to the last
committed offset on the relevant topic partitions. The consumer pause in
doSeekRetriableTopicPartitions ensures that we don't get messages for the
partitions that don't have failed tuples. maxPollRecords then caps how far past
the commit offset we can read on those partitions. If the same tuples are
always failing, we'll never get more than maxPollRecords from the commit offset
on those partitions as far as I can tell.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
