How does kafka consumer auto commit work?

问题:

I am reading this one:

Automatic Commit The easiest way to commit offsets is to allow the consumer to do it for you. If you configure enable.auto.commit=true, then every five seconds the consumer will commit the largest offset your client received from poll(). The five-second interval is the default and is controlled by setting auto.commit.interval.ms. Just like everything else in the consumer, the automatic commits are driven by the poll loop. Whenever you poll, the consumer checks if it is time to commit, and if it is, it will commit the offsets it returned in the last poll.

Maybe issue that my English is not good but I do not fully understood this description.

Let's say I use auto-commit with default interval - 5 sec and poll happens every 7 sec. At this case, a commit will happen every 5 sec or every 7 sec?

Can you clarify behaviour if poll will happen every 3 sec? Will commit happen every 5 sec or every 6 sec?
I have read this one:

Auto commits: You can set auto.commit to true and set the auto.commit.interval.ms property with a value in milliseconds. Once you've enabled this, the Kafka consumer will commit the offset of the last message received in response to its poll() call. The poll() call is issued in the background at the set auto.commit.interval.ms.

And it contradict the answer.

Can you explain this stuff in details.

let say I have diagramm like this:

0 sec - poll
4 sec - poll
8 sec - poll

When does offset will be committed and when which one?

回答1:

The auto-commit check is called in every poll and it checks that the time elapsed is greater than the configured time. If so, the offset is committed.

In case the commit interval is 5 seconds and poll is happening in 7 seconds, the commit will happen after 7 seconds only.

回答2:

It would try to autocommit ASAP after poll completes. You can have a look on the source code of consumer coordinator, which has set of local fields defined on class level to understand whether autocommit is enabled, what is the interval, and what is the next deadline to perform autocommit.

https://github.com/apache/kafka/blob/10cd98cc894b88c5d1e24fc54c66361ad9914df2/clients/src/main/java/org/apache/kafka/clients/consumer/internals/ConsumerCoordinator.java#L625

And one of the places within poll that perform a call to do storage https://github.com/apache/kafka/blob/10cd98cc894b88c5d1e24fc54c66361ad9914df2/clients/src/main/java/org/apache/kafka/clients/consumer/internals/ConsumerCoordinator.java#L279

That being said for instance poll executed every 7 seconds, and autocommit set to 5:

0 - poll, + set deadline to 5th second

7 - poll + commit due to deadline, update deadline to 7+5=12

14 - poll + commit due to deadline, update deadline to 12+5=17

However if polling set to every 3 seconds, and autocommit is set to 5:

0 - poll, + set deadline to 5th second

3 - poll, no commit

6 - poll + commit due to deadline, update deadline to 6+5=11