[
https://issues.apache.org/jira/browse/FLINK-20724?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Piotr Nowojski updated FLINK-20724:
-----------------------------------
Description:
This is an optimisation idea.
Create a similar http handler to {{AggregatingSubtasksMetricsHandler}}, but one
that would aggregate metrics per task, from all of the job vertices. The new
handler would only take {{JobID}} as a parameter. So that Web UI can in one RPC
obtain {{max(isBackPressureRatio)}} / {{max(isCausingBackPressureRatio)}} per
each task in the job graph.
This is related to FLINK-14712, where we are invoking more REST calls to get
the statistics per each task/node (in order to color the nodes based on the
back pressure and busy times). With this new handler, WebUI could make a one
REST call to get all the metrics it needs, instead of doing one REST call per
every Task.
was:
Create a similar http handler to {{AggregatingSubtasksMetricsHandler}}, but one
that would aggregate metrics per task, from all of the job vertices. The new
handler would only take {{JobID}} as a parameter. So that Web UI can in one RPC
obtain {{max(isBackPressureRatio)}} / {{max(isCausingBackPressureRatio)}} per
each task in the job graph.
This is related to FLINK-14712, where
> Create a http handler for aggregating metrics from whole job
> ------------------------------------------------------------
>
> Key: FLINK-20724
> URL: https://issues.apache.org/jira/browse/FLINK-20724
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Metrics
> Reporter: Piotr Nowojski
> Priority: Major
> Fix For: 1.13.0
>
>
> This is an optimisation idea.
> Create a similar http handler to {{AggregatingSubtasksMetricsHandler}}, but
> one that would aggregate metrics per task, from all of the job vertices. The
> new handler would only take {{JobID}} as a parameter. So that Web UI can in
> one RPC obtain {{max(isBackPressureRatio)}} /
> {{max(isCausingBackPressureRatio)}} per each task in the job graph.
> This is related to FLINK-14712, where we are invoking more REST calls to get
> the statistics per each task/node (in order to color the nodes based on the
> back pressure and busy times). With this new handler, WebUI could make a one
> REST call to get all the metrics it needs, instead of doing one REST call per
> every Task.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
