[
https://issues.apache.org/jira/browse/FLINK-13101?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Biao Liu updated FLINK-13101:
-----------------------------
Description:
The property blockingConnectionsBetweenChains means, if there are some stream
edges that can not be chained and the shuffle mode of edge is not specified,
translate these edges into {{BLOCKING}} result partition type.
The reason of introducing it is to satisfy the requirement of Blink batch
planner. Because the current scheduling strategy is a bit simple. It can not
support some complex scenarios, like a batch job with resources limited.
To be honest, it's probably a work-around solution. However it's an internal
implementation, we can replace it when we are able to support batch job by
scheduling strategy.
was:
The property "blocking after chaining off" means, if there are some stream
edges that can not be chained and the shuffle mode of edge is not specified,
translate these edges into {{BLOCKING}} result partition type.
The reason of introducing it is to satisfy the requirement of Blink batch
planner. Because the current scheduling strategy is a bit simple. It can not
support some complex scenarios, like a batch job with resources limited.
To be honest, it's probably a work-around solution. However it's an internal
implementation, we can replace it when we are able to support batch job by
scheduling strategy.
> Introduce blockingConnectionsBetweenChains property of StreamGraph
> ------------------------------------------------------------------
>
> Key: FLINK-13101
> URL: https://issues.apache.org/jira/browse/FLINK-13101
> Project: Flink
> Issue Type: Improvement
> Components: API / DataStream
> Reporter: Biao Liu
> Assignee: Biao Liu
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.9.0
>
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> The property blockingConnectionsBetweenChains means, if there are some stream
> edges that can not be chained and the shuffle mode of edge is not specified,
> translate these edges into {{BLOCKING}} result partition type.
> The reason of introducing it is to satisfy the requirement of Blink batch
> planner. Because the current scheduling strategy is a bit simple. It can not
> support some complex scenarios, like a batch job with resources limited.
> To be honest, it's probably a work-around solution. However it's an internal
> implementation, we can replace it when we are able to support batch job by
> scheduling strategy.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
