[ https://issues.apache.org/jira/browse/FLINK-6988?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16092701#comment-16092701 ]
ASF GitHub Bot commented on FLINK-6988: --------------------------------------- Github user tzulitai commented on the issue: https://github.com/apache/flink/pull/4239 Thanks a lot for opening a pull request for this very important feature @pnowojski. I did a rough first pass and had some comments I would like to clear out first (this is a big chunk of code, we would probably need to go through this quite a few times before it can be mergeable.) Most notably, some comments so far: 1. I think we need UTs for the `TwoPhaseCommitSinkFunction`. It alone is a very important addition (I would even prefer a separate PR for it and try to merge that first.) 2. Serialization of the transaction state in `TwoPhaseCommitSinkFunction` needs to be changed 2. Is the `FlinkKafkaProducer011` actually supporting hybrid (normal sink function and `writeToKafkaWithTimestamps` as a custom sink operator)? From the looks of it, it doesn't seem like it. > Add Apache Kafka 0.11 connector > ------------------------------- > > Key: FLINK-6988 > URL: https://issues.apache.org/jira/browse/FLINK-6988 > Project: Flink > Issue Type: Improvement > Components: Kafka Connector > Affects Versions: 1.3.1 > Reporter: Piotr Nowojski > Assignee: Piotr Nowojski > > Kafka 0.11 (it will be released very soon) add supports for transactions. > Thanks to that, Flink might be able to implement Kafka sink supporting > "exactly-once" semantic. API changes and whole transactions support is > described in > [KIP-98|https://cwiki.apache.org/confluence/display/KAFKA/KIP-98+-+Exactly+Once+Delivery+and+Transactional+Messaging]. > The goal is to mimic implementation of existing BucketingSink. New > FlinkKafkaProducer011 would > * upon creation begin transaction, store transaction identifiers into the > state and would write all incoming data to an output Kafka topic using that > transaction > * on `snapshotState` call, it would flush the data and write in state > information that current transaction is pending to be committed > * on `notifyCheckpointComplete` we would commit this pending transaction > * in case of crash between `snapshotState` and `notifyCheckpointComplete` we > either abort this pending transaction (if not every participant successfully > saved the snapshot) or restore and commit it. -- This message was sent by Atlassian JIRA (v6.4.14#64029)