spring boot - Why do I get two KubernetesClient beans only when running a native image? - Stack Overflow

admin管理员组

文章数量:1394593

I have a Spring Boot + Spring Cloud application that I'd like to build as a native image for faster startup time and smaller image size. I understand that Spring Cloud is not yet providing official support for native images, so I'm manually registering hints for reflection using Spring's RuntimeHintsRegistrar during AOT processing.

Now I'm able to build a native image with gradle bootBuildImage, but when I deploy the image to Kubernetes it fails with this error:

***************************
APPLICATION FAILED TO START
***************************

Description:

Parameter 0 of method kubernetesPodUtils in .springframework.cloud.kubernetes.fabric8.Fabric8AutoConfiguration required a single bean, but 2 were fou
    - kubernetesClient: defined in unknown location
    - configKubernetesClient: a programmatically registered singleton

The kubernetesClient bean is provided in Fabric8AutoConfiguration, while configKubernetesClient is registered programmatically in Fabric8ConfigDataLocationResolver.

This does not happen if I deploy a non-native image of the same exact application, so I guess it's the AOT processing that is missing something.

The app depends on spring-cloud-starter-kubernetes-fabric8-config and has the following properties set:

spring.config.import=kubernetes:
spring.cloud.kubernetes.config.reload.enabled=true

Why does this happen in the native image? Is the AOT processing missing some hint that I can register manually? I know I can workaround this by providing yet another KubernetesClient bean in my app code and annotate it with @Primary, but I'd prefer to use the one provided by Spring Cloud and not override it.

本文标签： spring bootWhy do I get two KubernetesClient beans only when running a native imageStack Overflow