[
https://issues.apache.org/jira/browse/FLINK-16733?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17084479#comment-17084479
]
Zili Chen edited comment on FLINK-16733 at 4/16/20, 2:05 AM:
-------------------------------------------------------------
[~rongr] I didn't communicate with [~xintongsong] yet but personally, it would
be benefit from how Kubernetes codepath works. That is,
1. We have a {{FlinkYARNClient}} handles jobs such as create AppContext,
ContainerLaunchContext and retrieve jobs.
2. For respect configuration, we have a decorator chain to decorate AppContext
& ContainerLaunchContext, but possibly shipping files are side-effect actions.
3. Adapt {{FlinkYARNClient}} to {{YARNClusterDescriptor}}, and supersede
{{Utils}}.
Generally speaking it is not something new but helps a lot we separate
concept(instead of putting messy stuff in {{YARNClusterDescriptor}} and
{{Utils}}), handle a specific thing in one pass(like step decorator, setup
launch command/ship files/config classpath/...) and gather related logic.
was (Author: tison):
[~rongr] I didn't communicate with [~xintongsong] yet but personally, it would
be benefit from how Kubernetes codepath works. That is,
1. We have a {{FlinkYARNClient}} handles jobs such as create AppContext,
ContainerLaunchContext and retrieve jobs.
2. For respect configuration, we have a decorator chain to decorate AppContext
& ContainerLaunchContext, but possibly shipping files are side-effect actions.
3. Adapt {{FlinkYARNClient}} to {{YARNClusterDescriptor}}, and supersede
{{Utils}}.
Generally speaking it is not something new but helps a lot we separate
concept(instead of putting messy stuff in {{YARNClusterDescriptor}} and
{{Utils}}), handle a specific thing in one pass and gather related logic.
> Refactor YarnClusterDescriptor
> ------------------------------
>
> Key: FLINK-16733
> URL: https://issues.apache.org/jira/browse/FLINK-16733
> Project: Flink
> Issue Type: Improvement
> Components: Deployment / YARN
> Reporter: Xintong Song
> Assignee: Xintong Song
> Priority: Minor
>
> Currently, YarnClusterDescriptor is not in a good shape. It has 1600+ lines
> of codes, of which the method {{startAppMaster}} alone has 400+ codes,
> leading to poor maintainability.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
