[
https://issues.apache.org/jira/browse/FLINK-26136?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17497185#comment-17497185
]
Gyula Fora commented on FLINK-26136:
------------------------------------
[~aitozi] the reason why need validation in both places is because not every
user will run the Webhook. In some environments the webhook cannot be deployed
so we can only have to operator.
In those case we need to make sure we run the validation in the reconcile loop
to and fatal errors should lead to an immediate ERROR status on the deployment.
And the ticket is for actually using the same logic in both places to be
consistent and simple
> Implement shared validation logic for FlinkDeployment objects
> -------------------------------------------------------------
>
> Key: FLINK-26136
> URL: https://issues.apache.org/jira/browse/FLINK-26136
> Project: Flink
> Issue Type: Sub-task
> Components: Kubernetes Operator
> Reporter: Gyula Fora
> Assignee: Gyula Fora
> Priority: Major
>
> At the moment there is only a very basic “placeholder” validation logic
> implemented in the webhook module:
> org.apache.flink.kubernetes.operator.admission.FlinkDeploymentValidator
> We should aim to validate parts of the FlinkDeployment that can be done
> upfront, things like most common Flink config options, parallelism, resources
> etc.
> As described in https://issues.apache.org/jira/browse/FLINK-26135 this
> validation should be part of the flink-kubernetes-operator module.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
