Timo Theusner created FLINK-41213:
-------------------------------------
Summary: Submit refresh jobs from programs deployed in application
mode
Key: FLINK-41213
URL: https://issues.apache.org/jira/browse/FLINK-41213
Project: Flink
Issue Type: Sub-task
Components: Table SQL / Planner, Client / Job Submission
Reporter: Timo Theusner
Read {{execution.target}} from the {{TableConfig}} when set there, ignoring the
root configuration's cluster-selecting options, and submit through that
target's pipeline executor.
Proposed Changes
* Resolve the refresh target from the {{TableConfig}} if it sets
{{{}execution.target{}}}, else from the root configuration, for submission and
job control
* Strip {{{}rest.address{}}}, {{{}jobmanager.rpc.address{}}},
{{{}high-availability.*{}}}, {{kubernetes.cluster-id}} and
{{yarn.application.id}} from the root layer when the {{TableConfig}} sets the
target ({{{}RefreshJobConfigurationLayers{}}})
* Reject a {{TableConfig}} target without its cluster ID on the
{{{}TableConfig{}}}, e.g. {{remote}} without {{{}rest.address{}}}, before any
catalog change
* Submit refresh jobs through the {{PipelineExecutorFactory}} of the refresh
target instead of the context environment's executor
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
