[
https://issues.apache.org/jira/browse/FLINK-24257?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17414807#comment-17414807
]
Timo Walther commented on FLINK-24257:
--------------------------------------
[~zhihao] thanks for sharing these ideas with us. such a proposal should rather
be discussed on the dev@ ML using the {{[DISCUSS]}} tag to have more exposure.
In general, I think similar approaches have been implemented already internally
at users with larger deployments.
> Using Stateful Blue/Green Deployment to handle Application Evolvement
> ---------------------------------------------------------------------
>
> Key: FLINK-24257
> URL: https://issues.apache.org/jira/browse/FLINK-24257
> Project: Flink
> Issue Type: Improvement
> Reporter: Zhihao Wang
> Priority: Major
>
> Application Evolvement is an open problem in streaming processing. We present
> a *stateful blue/green deployment* to address it. To our knowledge, this is
> the first general way to handle streaming application evolvement.
> Here is the design doc:
> [https://docs.google.com/document/d/1QCnhtrgcMRUWCw_SO7vmfj6dyzbdjIAxWkJzKK4d508|https://docs.google.com/document/d/1QCnhtrgcMRUWCw_SO7vmfj6dyzbdjIAxWkJzKK4d508/edit?usp=sharing]
>
> The design requires a backfill requirement supported by Flink. Please help to
> review the design and to see Flink community can support such a feature.
>
> P.S. [A related ticket|https://issues.apache.org/jira/browse/KAFKA-13291] is
> also sent to Kafka community for suggestions.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
