[
https://issues.apache.org/jira/browse/FLINK-19937?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17228952#comment-17228952
]
Lsw_aka_laplace commented on FLINK-19937:
-----------------------------------------
[~hackergin]
Hi jinfeng,
I have similar concern with yours, especially when I applied SINK_PARALLELISM
to specific connectors. In `PhysicalCommonSink`, for now, we can set the
parallelism but there is no access to setting the partitioning stretagy.
Sometimes it can be misleading or may cause some unexpected condition.
Well, the concern mentioned above seems another theme. From my perspective,
there are basically two ways.
1. Just keep it, meanwhile let users know the concern if they want to
configure the sink parallelism.
2. Add a new Interface/ add a new method in `ParallelismProvider` called
`paratitioningStrategy` to give the access so that users can choose their own
strategy.
Well, both ways require user‘s knowledgement on DataStream, which seems to be
the contrary to 'Less Knowledgement'.
> Support sink parallelism option for all connectors
> --------------------------------------------------
>
> Key: FLINK-19937
> URL: https://issues.apache.org/jira/browse/FLINK-19937
> Project: Flink
> Issue Type: New Feature
> Components: Table SQL / Ecosystem
> Reporter: Lsw_aka_laplace
> Priority: Major
> Labels: pull-request-available
>
> Since https://issues.apache.org/jira/browse/FLINK-19727 has been done.
> SINK_PARALLELISM option and `ParallelismProvider` should be applied for all
> existing `DynamicTableSink` of connectors in order to give users access to
> setting their own sink parallelism.
>
>
> Update：
> Anybody who works on this issue should refrence to FLINK-19727~
> `ParallelismProvider` should work with `SinkRuntimeProvider`, actually
> `SinkFunctionProvider` and `OutputFormatProvider` has implemented
> `ParallelismProvider`. And `SINK_PARALLELISM` has already defined in
> `FactoryUtil`, plz reuse it.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
