[
https://issues.apache.org/jira/browse/FLINK-12607?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17286279#comment-17286279
]
John Phelan edited comment on FLINK-12607 at 2/18/21, 5:33 AM:
---------------------------------------------------------------
hi [~rmetzger] I made what I think is a first pass at the changes:
[[FLINK-12607] working on REST api returns max parallelism ·
bytesandwich/flink@6773df8
(github.com)|https://github.com/bytesandwich/flink/commit/1c63556815c3eb661147b685fa0d18c487b6cf2a]
I just wanted to circle back and see if I'm making a reasonable approach and to
ask what testing would make sense.
h1. Questions:
It's possible to set maxParallelism in a global way, and it's also possible to
set it on each individual operator? The maxparallelism that's set on the
execution environment seems to be what should get returned in the
|h5. */jobs/:jobid*|
is that correct? To distinguish it from the internal max-parallelisms, I called
this:
|h5. *job-max-parallelism*|
We seem to be reading in a serializable ArchivedExecutionConfig at some point,
which I think would read in the default maxParallelism of 0 if it reads an
object serialized before this change and I could see this being problematic. I
bumped the serialVersionUID just to call attention to this.
was (Author: bytesandwich):
hi [~rmetzger] I made what I think is a first pass at the changes:
[[FLINK-12607] working on REST api returns max parallelism ·
bytesandwich/flink@6773df8
(github.com)|https://github.com/bytesandwich/flink/commit/6773df87acdf862718374a2e31dd0d03b90827f7]
I just wanted to circle back and see if I'm making a reasonable approach and to
ask what testing would make sense.
h1. Questions:
It's possible to set maxParallelism in a global way, and it's also possible to
set it on each individual operator? The maxparallelism that's set on the
execution environment seems to be what should get returned in the
|h5. */jobs/:jobid*|
is that correct? To distinguish it from the internal max-parallelisms, I called
this:
|h5. *job-max-parallelism*|
We seem to be reading in a serializable ArchivedExecutionConfig at some point,
which I think would read in the default maxParallelism of 0 if it reads an
object serialized before this change and I could see this being problematic. I
bumped the serialVersionUID just to call attention to this.
> Introduce a REST API that returns the maxParallelism of a job
> -------------------------------------------------------------
>
> Key: FLINK-12607
> URL: https://issues.apache.org/jira/browse/FLINK-12607
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / REST
> Affects Versions: 1.6.3
> Reporter: Akshay Kanfade
> Assignee: John Phelan
> Priority: Minor
> Labels: starter
>
> Today, Flink does not offer any way to get the maxParallelism for a job and
> it's operators through any of the REST APIs. Since, the internal state
> already tracks maxParallelism for a job and it's operators, we should expose
> this via the REST APIs so that application developer can get more insights on
> the current state.
> There can be two approaches on how we can do this -
> Approach 1 :
> Modify the existing rest API response model to additionally expose a new
> field 'maxParallelism'. Some of the REST APIs that would be affected by this
> |h5. */jobs/:jobid/vertices/:vertexid*|
> |h5. */jobs/:jobid*|
>
> Approach 2 :
> Create a new REST API that would only return maxParallelism for a job and
> it's operators.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
