You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "ASF GitHub Bot (Jira)" <ji...@apache.org> on 2022/08/20 12:33:00 UTC

[jira] [Updated] (FLINK-28831) [3] Support pluginable decorators mechanism

     [ https://issues.apache.org/jira/browse/FLINK-28831?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

ASF GitHub Bot updated FLINK-28831:
-----------------------------------
    Labels: pull-request-available  (was: )

> [3] Support pluginable decorators mechanism
> -------------------------------------------
>
>                 Key: FLINK-28831
>                 URL: https://issues.apache.org/jira/browse/FLINK-28831
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Deployment / Kubernetes
>            Reporter: bo zhao
>            Priority: Major
>              Labels: pull-request-available
>
> Introduce a pluginable decorators mechanism into Flink JobManager and TaskManager. Currently, all decorators are hard-coded in the JobManager deployment and TaskManager pod K8S creation.
>  
> We propose using SPI or Flink plugins mechanism to load the external decorators. In this way, we can make Flink more flexible towards supporting the decorators of customized K8S schedulers.
> We propose a new plugin mechanism for supporting load the jar package of a single customized scheduler. The driver jar package should contain the specific K8sStepDecorator and its dependencies. Taking an example, the jar package should be packaged into the ‘opt’ directory, and contains 2 major things:
>  # A K8sStepDecorator implemented by Customized K8S SchedulerA.
>  # The all dependencies from the introduced K8sStepDecorator.
> when users want to use the customized scheduler A, he/she need to create a new directory which named as the customized scheduler’s name A, then copy & paste the said jar into the new directory, that could be loaded when user specific the related Flink K8S configuration options and enable the functionality in Flink.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)