Github user srdo commented on the issue:
https://github.com/apache/storm/pull/1919
Autocommit=true + ack=0 means the consumer will commit periodically. Tuples
cannot fail. Messages may be replayed if the spout crashes (it may not have
committed them because it only happens periodically). This is not true
at-most-once, since it is only at-most-once if the spout doesn't crash.
Autocommit=true + ack>0. Acked tuples are handled identically to the case
above. Failed tuples are replayed, but only if the spout doesn't crash. This
configuration could make sense if you just want best-effort replaying, or
you're running a topology with other types of spouts as well.
Autocommit=false + ack=0 will cause Storm to ack tuples immediately. The
spout will commit them periodically
(https://github.com/XuMingmin/storm/blob/717edc8e7ea46ef3392ec1402235a764039070ff/external/storm-kafka-client/src/main/java/org/apache/storm/kafka/spout/KafkaSpout.java#L241).
This is identical behavior to the first case, except the commit handling has
moved from the consumer to the spout.
Autocommit=false + ack>0 is at-least-once.
---
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.
---
