[
https://issues.apache.org/jira/browse/KAFKA-10395?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17180881#comment-17180881
]
Matthias J. Sax commented on KAFKA-10395:
-----------------------------------------
[~ableegoldman] [~vvcephei] Seems the PR was merged. Can we close the ticket?
> TopologyTestDriver does not work with dynamic topic routing
> -----------------------------------------------------------
>
> Key: KAFKA-10395
> URL: https://issues.apache.org/jira/browse/KAFKA-10395
> Project: Kafka
> Issue Type: Bug
> Components: streams
> Reporter: Sophie Blee-Goldman
> Assignee: Sophie Blee-Goldman
> Priority: Major
> Labels: test-framework
> Fix For: 2.7.0
>
>
> The TopologyTestDriver#read(topic) methods all call #getRecordsQueue which
> checks
>
> {code:java}
> final Queue<ProducerRecord<byte[], byte[]>> outputRecords =
> outputRecordsByTopic.get(topicName);
> if (outputRecords == null) {
> if (!processorTopology.sinkTopics().contains(topicName)) {
> throw new IllegalArgumentException("Unknown topic: " + topicName);
> }
> }
> {code}
> The outputRecordsByTopic map keeps track of all topics that are actually
> produced to, but obviously doesn't capture any topics that haven't yet
> received output. The `processorTopology#sinkTopics` is supposed to account
> for that by checking to make sure the topic is actually registered in the
> topology, and throw an exception if not in case the user supplied the wrong
> topic name to read from.
> Unfortunately the TopicNameExtractor allows for dynamic routing of records to
> any topic, so the topology isn't aware of all the possible output topics. If
> trying to read from one of these topics that happens to not have received any
> output yet, the test will throw the above misleading IllegalArgumentException.
> We could just relax this check, but warning users who may actually have
> accidentally passed in the wrong topic to read from seems quite useful. A
> better solution would be to require registering all possible output topics to
> the TTD up front. This would obviously require a KIP, but it would be a very
> small one and shouldn't be too much trouble
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
