You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "Fede (Jira)" <ji...@apache.org> on 2024/02/22 20:52:00 UTC
[jira] [Updated] (KAFKA-16300) Wrong documentation for producer config retries
[ https://issues.apache.org/jira/browse/KAFKA-16300?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Fede updated KAFKA-16300:
-------------------------
Description:
In documentation from version 3.1 to version 3.4, it looks like the retries explanation has a bug related to max.in.flight.request.per.connection related parameter and possible message reordering.
[https://kafka.apache.org/31/documentation.html#producerconfigs_retries]
[https://kafka.apache.org/32/documentation.html#producerconfigs_retries]
[https://kafka.apache.org/33/documentation.html#producerconfigs_retries]
[https://kafka.apache.org/34/documentation.html#producerconfigs_retries]
in particular, the section
Allowing retries while setting enable.idempotence to false and max.in.flight.requests.per.connection to 1 will potentially change the ordering of records because if two batches are sent to a single partition, and the first fails and is retried but the second succeeds, then the records in the second batch may appear first.
Is states
max.in.flight.requests.per.connection to 1
It should be said
max.in.flight.requests.per.connection to *greater than* 1
This bug has been fixed in the latest versions, but it still confuses users using affected versions as the meaning is the opposite of what it should be.
I created a PR ([https://github.com/apache/kafka/pull/15413|https://github.com/apache/kafka/pull/15413)])for version 3.2, but the build failed. Not sure why.
was:
In documentation from version 3.1 to version 3.4, it looks like the retries explanation has a bug related to max.in.flight.request.per.connection related parameter and possible message reordering.
[https://kafka.apache.org/31/documentation.html#producerconfigs_retries]
[https://kafka.apache.org/32/documentation.html#producerconfigs_retries]
[https://kafka.apache.org/33/documentation.html#producerconfigs_retries]
[https://kafka.apache.org/34/documentation.html#producerconfigs_retries]
in particular, the section
Allowing retries while setting enable.idempotence to false and max.in.flight.requests.per.connection to 1 will potentially change the ordering of records because if two batches are sent to a single partition, and the first fails and is retried but the second succeeds, then the records in the second batch may appear first.
Is states
max.in.flight.requests.per.connection to 1
It should be said
max.in.flight.requests.per.connection to *greater than* 1
This bug has been fixed in the latest versions, but it still confuses users using affected versions as the meaning is the opposite of what it should be.
I created a PR ([https://github.com/apache/kafka/pull/15413)] for version 3.2, but the build failed. Not sure why.
> Wrong documentation for producer config retries
> -----------------------------------------------
>
> Key: KAFKA-16300
> URL: https://issues.apache.org/jira/browse/KAFKA-16300
> Project: Kafka
> Issue Type: Bug
> Components: documentation
> Affects Versions: 3.1.2, 3.3, 3.2.3, 3.4.1
> Reporter: Fede
> Priority: Minor
> Labels: documentation
>
> In documentation from version 3.1 to version 3.4, it looks like the retries explanation has a bug related to max.in.flight.request.per.connection related parameter and possible message reordering.
> [https://kafka.apache.org/31/documentation.html#producerconfigs_retries]
> [https://kafka.apache.org/32/documentation.html#producerconfigs_retries]
> [https://kafka.apache.org/33/documentation.html#producerconfigs_retries]
> [https://kafka.apache.org/34/documentation.html#producerconfigs_retries]
>
> in particular, the section
> Allowing retries while setting enable.idempotence to false and max.in.flight.requests.per.connection to 1 will potentially change the ordering of records because if two batches are sent to a single partition, and the first fails and is retried but the second succeeds, then the records in the second batch may appear first.
>
> Is states
> max.in.flight.requests.per.connection to 1
>
> It should be said
> max.in.flight.requests.per.connection to *greater than* 1
>
> This bug has been fixed in the latest versions, but it still confuses users using affected versions as the meaning is the opposite of what it should be.
>
> I created a PR ([https://github.com/apache/kafka/pull/15413|https://github.com/apache/kafka/pull/15413)])for version 3.2, but the build failed. Not sure why.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)