[
https://issues.apache.org/jira/browse/FLINK-5133?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15882693#comment-15882693
]
ASF GitHub Bot commented on FLINK-5133:
---------------------------------------
Github user StephanEwen commented on the issue:
https://github.com/apache/flink/pull/3303
A more general question on the resource matching: If I understand it
correctly, then the resource manager will try to get the "max" resources for an
operator, but potentially go down to the "min" resources if it cannot get the
"max" resources.
I am wondering if "max" is confusing here and we should rather call it
"preferred". Calling it "max" would make users think that those are the maximum
resources that the operator can handle. I think many users would set max very
high, thinking that this gives Flink more freedom to find suitable resources.
They probably do not expect from the name that it always tries to satisfy the
"max" resources.
> Support to set resource for operator in DataStream and DataSet
> --------------------------------------------------------------
>
> Key: FLINK-5133
> URL: https://issues.apache.org/jira/browse/FLINK-5133
> Project: Flink
> Issue Type: Sub-task
> Components: DataSet API, DataStream API
> Reporter: zhijiang
> Assignee: zhijiang
>
> This is part of the fine-grained resource configuration.
> For *DataStream*, the *setResource* API will be setted onto
> *SingleOutputStreamOperator* similar with other existing properties like
> parallelism, name, etc.
> For *DataSet*, the *setResource* API will be setted onto *Operator* in the
> similar way.
> There are two parameters described with minimum *ResourceSpec* and maximum
> *ResourceSpec* separately in the API for considering resource resize in
> future improvements.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
