[
https://issues.apache.org/jira/browse/FLINK-32488?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17763992#comment-17763992
]
Yangze Guo commented on FLINK-32488:
------------------------------------
Thanks for the PR [~hejufang001] and [~xiangyu0xf] . I'm now considering not
introducing a configuration for it. As we all know, the configuration of flink
become massive which would harm the usability. How about make the refresh
interval adaptive to the total number of execution cache in job manager?
> Introduce configuration to control ExecutionGraph cache in REST API
> -------------------------------------------------------------------
>
> Key: FLINK-32488
> URL: https://issues.apache.org/jira/browse/FLINK-32488
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / REST
> Affects Versions: 1.16.2, 1.17.1
> Reporter: Hong Liang Teoh
> Assignee: Jufang He
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.18.0
>
>
> *What*
> Currently, REST handlers that inherit from AbstractExecutionGraphHandler
> serve information derived from a cached ExecutionGraph.
> This ExecutionGraph cache currently derives it's timeout from
> {*}web.refresh-interval{*}. The *web.refresh-interval* controls both the
> refresh rate of the Flink dashboard and the ExecutionGraph cache timeout.
> We should introduce a new configuration to control the ExecutionGraph cache,
> namely {*}rest.cache.execution-graph.expiry{*}.
> *Why*
> Sharing configuration between REST handler and Flink dashboard is a sign that
> we are coupling the two.
> Ideally, we want our REST API behaviour to independent of the Flink dashboard
> (e.g. supports programmatic access).
>
> Mailing list discussion:
> https://lists.apache.org/thread/7o330hfyoqqkkrfhtvz3kp448jcspjrm
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
