[SPARK-36366] Google Kubernetes Engine authentication fails

• Type: Bug
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 3.1.2
• Component/s: Kubernetes
• Labels: google kubernetes kubernetesexecutor newbie

Environment

$ kubectl version
Client Version: version.Info{Major:"1", Minor:"18", GitVersion:"v1.18.15", GitCommit:"73dd5c840662bb066a146d0871216333181f4b64", GitTreeState:"clean", BuildDate:"2021-01-13T13:22:41Z", GoVersion:"go1.13.15", Compiler:"gc", Platform:"darwin/amd64"}
Server Version: version.Info{Major:"1", Minor:"18+", GitVersion:"v1.18.19-gke.1701", GitCommit:"d7cecefb99b58e8968f59b59d76448eb1e6ea403", GitTreeState:"clean", BuildDate:"2021-06-23T21:51:59Z", GoVersion:"go1.13.15b4", Compiler:"gc", Platform:"linux/amd64"}

$ spark-submit --version
version 3.1.2
Using Scala version 2.12.10, OpenJDK 64-Bit Server VM, 11.0.10

• Assignee: Unassigned
• Reporter: Tiago Reis
• Shepherd: Yinan Li
• Created: 30/Jul/21 22:53
• Updated: 30/Jul/21 22:53

Description

When connecting to a Google Kubernetes Engine, a command gcloud container clusters get-credentials is used that generates a ~/.kube/config file. The distinctive trait in this config file is that it uses an auth-provider relying on gcloud to inject the keys expiry and access-token from the general Google SDK auth config, as seen here:

users:
- name: gke_my-project_my-region_my-cluster
  user:
    auth-provider:
      config:
        cmd-args: config config-helper --format=json
        cmd-path: /Users/user/google-cloud-sdk/bin/gcloud
        expiry-key: '{.credential.token_expiry}'
        token-key: '{.credential.access_token}'

kubectl, because it uses client-go, supports the auth-provider and fetches the token and expiry from the json returne by config-helper. As Spark is using the fabric8 client, this is yet to be supported, breaking when running spark-submit:

Exception in thread "main" io.fabric8.kubernetes.client.KubernetesClientException: Failure executing: POST at: https://my-endpoint/api/v1/namespaces/my-namespace/pods. Message: Forbidden! User gke_my-project_my-region_my-cluster doesn't have permission. pods is forbidden: User "system:anonymous" cannot create resource "pods" in API group "" in the namespace "my-namespace".