[
https://issues.apache.org/jira/browse/FLINK-8949?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16400326#comment-16400326
]
Chesnay Schepler commented on FLINK-8949:
-----------------------------------------
huh, well that's a problem.
Unfortunately the UI actually requires the watermark for each subtask,
otherwise we could've just queried the aggregated minimum instead.
The metric querying was always a bit verbose but I don't see us reworking that
anytime soon; although with the aggregating handlers we conceptually know how
to do it,
I guess we'll just have to batch the requests in multiples of 50(?) or
something.
> Rest API failure with long URL
> ------------------------------
>
> Key: FLINK-8949
> URL: https://issues.apache.org/jira/browse/FLINK-8949
> Project: Flink
> Issue Type: Bug
> Components: REST, Webfrontend
> Affects Versions: 1.5.0, 1.4.2
> Reporter: Truong Duc Kien
> Priority: Major
>
> When you have jobs with high parallelism, the URL for a REST request can get
> very long. When the URL is longer than 4096 bytes, the REST API will return
> error
> {{Failure: 404 Not Found}}
> This can easily be seen in the Web UI, when Flink queries for the watermark
> using the REST API:
> {{GET
> /jobs/:jobId/vertices/:vertexId/metrics?get=0.currentLowWatermark,1.currentLowWatermark,2.currentLo...}}
> The request will fail with more than 170 subtasks and the watermark will not
> be displayed in the Web UI.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
