[ 
https://issues.apache.org/jira/browse/FLINK-27029?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17517631#comment-17517631
 ] 

Gyula Fora edited comment on FLINK-27029 at 4/5/22 6:25 PM:
------------------------------------------------------------

What I mean is that we should merge the default config with the userProvided 
config before doing validation.

Example:

Default config:

high-availability: 
org.apache.flink.kubernetes.highavailability.KubernetesHaServicesFactory
high-availability.storageDir: [file:///flink-data/ha]

User config:

# config without ha settings

If the user enables last-state currently validation fails as no HA settings 
have been provided. This logic is broken because HA configs would actually come 
already from defaults. So we should not validate the user flinkConf directly, 
but first copy the default + apply user conf on top of it, then validate


was (Author: gyfora):
What I mean is that we should merge the default config with the userProvided 
config before doing validation.

Example:

Default config:

high-availability: 
org.apache.flink.kubernetes.highavailability.KubernetesHaServicesFactory
high-availability.storageDir: file:///flink-data/ha

User config:
# config without ha settings

If the user enables last-state currently validation fails as no HA settings 
have been provided. This logic is broken because HA configs would actually come 
already from defaults. So we should not validate the user flinkConf directly, 
but first copy the default + apply user conf on top of it, then validate

> DeploymentValidator should take default flink config into account during 
> validation
> -----------------------------------------------------------------------------------
>
>                 Key: FLINK-27029
>                 URL: https://issues.apache.org/jira/browse/FLINK-27029
>             Project: Flink
>          Issue Type: Bug
>          Components: Kubernetes Operator
>            Reporter: Gyula Fora
>            Priority: Major
>             Fix For: kubernetes-operator-1.0.0
>
>
> Currently the DefaultDeploymentValidator only takes the FlinkDeployment 
> object into account.
> However in places where we validate the presence of config keys we should 
> also consider the default flink config which might already provide default 
> values for the required configs even if the deployment itself doesnt.
> We should make sure this works correctly both in the operator and the webhook



--
This message was sent by Atlassian Jira
(v8.20.1#820001)

Reply via email to