[ https://issues.apache.org/jira/browse/BEAM-52?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15264860#comment-15264860 ]
Raghu Angadi commented on BEAM-52: ---------------------------------- Ah, thanks. just skimmed through it. It looks pretty much on the same lines as what I have. couple of differences : - Kafka producer uses the key for selecting a partition. I wanted to retain that functionality for users. So I apply our coders inside custom Kafka serializers. Otherwise Kafka will hash on serialized byte array. - I noticed you are catching the exceptions in a callback and reporting it back.. may be I should do that too. will get Dan's opinion as well in PR. I will ping you to review my pull request. > KafkaIO - bounded/unbounded, source/sink > ---------------------------------------- > > Key: BEAM-52 > URL: https://issues.apache.org/jira/browse/BEAM-52 > Project: Beam > Issue Type: New Feature > Components: sdk-java-extensions > Reporter: Daniel Halperin > Assignee: Raghu Angadi > > We should support Apache Kafka. The priority list is probably: > * UnboundedSource > * unbounded Sink > * BoundedSource > * bounded Sink > The connector should be well-tested, especially around UnboundedSource > checkpointing and resuming, and data duplication. -- This message was sent by Atlassian JIRA (v6.3.4#6332)