[ https://issues.apache.org/jira/browse/KAFKA-4875?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15902975#comment-15902975 ]
Michael Noll commented on KAFKA-4875: ------------------------------------- There two things at play here IMHO: 1. Whether or not there's a gap in the documentation (fixing this would be quick) 2. Whether or not the API/behavior should stay the way it is. I'm more concerned about this than about #1. > Kafka streams topic groups and builder.stream API > ------------------------------------------------- > > Key: KAFKA-4875 > URL: https://issues.apache.org/jira/browse/KAFKA-4875 > Project: Kafka > Issue Type: Improvement > Components: streams > Affects Versions: 0.10.2.0 > Reporter: Eno Thereska > Fix For: 0.11.0.0 > > > One thing that has come up in mailing list is that the notion of topic groups > is unclear. So if you have 2 topics, each with 3 partitions, you'd expect 6 > tasks. However, if you do `builder.stream(topic1, topic2)` you actually get > only 3 tasks created. If you do `builder.stream(topic1); > builder.stream(topic2)` you get 6 tasks, i.e., parallelism is increased. So > the same application, calling builder.stream() in two different ways, might > see different performance. > In the Kafka Streams documentations we mention partitions and tasks, but not > topic groups. We also do not document the effects of using builder.stream > with a topic array. We also need to revisit whether the API and its effects > are confusing. -- This message was sent by Atlassian JIRA (v6.3.15#6346)