[
https://issues.apache.org/jira/browse/FLINK-12691?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16865521#comment-16865521
]
vinoyang edited comment on FLINK-12691 at 6/17/19 11:11 AM:
------------------------------------------------------------
[~aljoscha] Thanks for sharing your idea. Based on your idea, users' records
need to be observed (by implementing the {{AsyncObserverFunction}}) that means
the record need to carry external information about the queue's capacity and
timeout? IMO, it's not a good design to extract the information from the users'
records. Two reasons:
* The record should not care about the detail of the engine;
* The frequency of the timeout been changed is very low;
was (Author: yanghua):
[~aljoscha] Thanks for sharing your idea. Based on your idea, users' records
need to be observed (by implementing the {{AsyncObserverFunction}}) that means
the record need to carry external information about the queue's capacity and
timeout? IMO, it's not a good design to extract the information from the users'
records.
> Make Queue Capacity and Timeout of AsyncWaitOperator changeable during runtime
> ------------------------------------------------------------------------------
>
> Key: FLINK-12691
> URL: https://issues.apache.org/jira/browse/FLINK-12691
> Project: Flink
> Issue Type: Improvement
> Components: API / DataStream
> Reporter: Konstantin Knauf
> Priority: Minor
>
> A user that I have recently been working with has the requirement to change
> the capacity and possibly the timeout during the runtime of a job.
> -Conceptually, this should be possibly by just passing these parameters to
> the {{RichAsyncFunction}} via its RuntimeContext. The change of the timeout
> would only apply to future records and the change in the queue capacity would
> take effect immediately.-
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
