mjsax commented on pull request #8504: URL: https://github.com/apache/kafka/pull/8504#issuecomment-622651059
> we should create two repartiton topics as that's what the user is expecting. Is she? I guess it's clearly document that the name is use for repartition topics, too, but it might still be an unexpected "side effect"? But I see the point that it's not too easy to resolve: Maybe we could log a warning about creating an unnecessary duplicated repartition topics (and update the docs how to resolve this, ie, via calling `repartition()` upfront before the join to specify one name)? Just picking one of both names seems wrong. Throwing an exception and refuse to build the topology seems too harsh? Overall, it seems it might be best to merge the PR using the current approach, and merge the repartition topics directly in `StreamsBuilder` instead of waiting for the optimization (this also avoid the issue with index bumping and is better default behavior anyway). Can you add the test for streamA-tableB, streamA-tableC join? ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
