You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@flink.apache.org by "Piotr Nowojski (JIRA)" <ji...@apache.org> on 2017/07/17 09:23:00 UTC

[jira] [Updated] (FLINK-7210) Add TwoPhaseCommitSinkFunction (implementing exactly-once semantic in generic way)

     [ https://issues.apache.org/jira/browse/FLINK-7210?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Piotr Nowojski updated FLINK-7210:
----------------------------------
    Description: To implement exactly-once sink there is a re-occurring pattern for doing it - two phase commit algorithm. It is used both in `BucketingSink` and in `Pravega` sink and it will be used in `Kafka 0.11` connector. It would be good to extract this common logic into one class, both to improve existing implementation (for exampe `Pravega`'s sink doesn't abort interrupted transactions) and to make it easier for the users to implement their own custom exactly-once sinks.  (was: To implement exactly-once sink there is a re-occurring pattern for doing it - two phase commit algorithm. It is used both in `BucketedSink` and in `Pravega` sink and it will be used in `Kafka 0.11` connector. It would be good to extract this common logic into one class, both to improve existing implementation (for exampe `Pravega`'s sink doesn't abort interrupted transactions) and to make it easier for the users to implement their own custom exactly-once sinks.)

> Add TwoPhaseCommitSinkFunction (implementing exactly-once semantic in generic way)
> ----------------------------------------------------------------------------------
>
>                 Key: FLINK-7210
>                 URL: https://issues.apache.org/jira/browse/FLINK-7210
>             Project: Flink
>          Issue Type: Improvement
>            Reporter: Piotr Nowojski
>            Assignee: Piotr Nowojski
>
> To implement exactly-once sink there is a re-occurring pattern for doing it - two phase commit algorithm. It is used both in `BucketingSink` and in `Pravega` sink and it will be used in `Kafka 0.11` connector. It would be good to extract this common logic into one class, both to improve existing implementation (for exampe `Pravega`'s sink doesn't abort interrupted transactions) and to make it easier for the users to implement their own custom exactly-once sinks.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)