Github user hmcl commented on a diff in the pull request:
https://github.com/apache/storm/pull/2537#discussion_r165851895
--- Diff:
external/storm-kafka-client/src/main/java/org/apache/storm/kafka/spout/KafkaSpoutConfig.java
---
@@ -143,28 +146,28 @@ public String toString() {
/**
* This enum controls when the tuple with the {@link ConsumerRecord}
for an offset is marked as processed,
- * i.e. when the offset can be committed to Kafka.
+ * i.e. when the offset can be committed to Kafka. The default value
is AT_LEAST_ONCE.
* The commit interval is controlled by {@link
KafkaSpoutConfig#getOffsetsCommitPeriodMs() }, if the mode commits on an
interval.
* NO_GUARANTEE may be removed in a later release without warning,
we're still evaluating whether it makes sense to keep.
- *
- * <ul>
- * <li>AT_LEAST_ONCE - an offset is ready to commit only after the
corresponding tuple has been processed and acked (at least once). If
- * a tuple fails or times out it will be re-emitted, as controlled by
the {@link KafkaSpoutRetryService}. Commits on the defined
- * interval.</li>
- * <br/>
- * <li>AT_MOST_ONCE - every offset will be committed to Kafka right
after being polled but before being emitted to the downstream
- * components of the topology. The commit interval is ignored. This
mode guarantees that the offset is processed at most once by
- * ensuring the spout won't retry tuples that fail or time out after
the commit to Kafka has been done.</li>
- * <br/>
- * <li>NO_GUARANTEE - the polled offsets are ready to commit
immediately after being polled. The offsets are committed periodically,
- * i.e. a message may be processed 0, 1 or more times. This behavior
is similar to setting enable.auto.commit=true in the consumer, but
- * allows the spout to control when commits occur. Commits on the
defined interval. </li>
- * </ul>
*/
@InterfaceStability.Unstable
public enum ProcessingGuarantee {
+ /**
+ * An offset is ready to commit only after the corresponding tuple
has been processed and acked (at least once). If a tuple fails or
+ * times out it will be re-emitted, as controlled by the {@link
KafkaSpoutRetryService}. Commits on the defined interval.
+ */
AT_LEAST_ONCE,
+ /**
+ * Every offset will be committed to Kafka right after being
polled but before being emitted to the downstream components of the
+ * topology. The commit interval is ignored. This mode guarantees
that the offset is processed at most once by ensuring the spout
+ * won't retry tuples that fail or time out after the commit to
Kafka has been done
+ */
AT_MOST_ONCE,
+ /**
+ * The polled offsets are ready to commit immediately after being
polled. The offsets are committed periodically, i.e. a message may
+ * be processed 0, 1 or more times. This behavior is similar to
setting enable.auto.commit=true in the consumer, but allows the
+ * spout to control when commits occur. Commits on the defined
interval
--- End diff --
Commits are made asynchronously on the defined interval.
Should we also say specifically that for A_L_O and A_M_O the commits are
done synchronously ?
---
