在AKS集群中实现外部Pod执行Notebook Pod内Python脚本的方法

To automate running your Python script inside the my-notebook-deployment Pod, you can create a Kubernetes Job (purpose-built for one-time, batch tasks) that uses kubectl to execute the script in your target Pod. Here's a step-by-step implementation:

1. Set Up RBAC Permissions #

First, we need to grant the Job's service account permission to run commands in the spark namespace.

Create a Service Account #

Save this as spark-executor-sa.yaml:

apiVersion: v1
kind: ServiceAccount
metadata:
  name: spark-script-executor
  namespace: spark

Create a Role with Exec Permissions #

Save this as pod-exec-role.yaml:

apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  name: pod-exec-role
  namespace: spark
rules:
- apiGroups: [""]
  resources: ["pods", "pods/exec"]
  verbs: ["get", "list", "create"]

Bind the Role to the Service Account #

Save this as spark-executor-rolebinding.yaml:

apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  name: spark-script-executor-binding
  namespace: spark
subjects:
- kind: ServiceAccount
  name: spark-script-executor
  namespace: spark
roleRef:
  kind: Role
  name: pod-exec-role
  apiGroup: rbac.authorization.k8s.io

Apply these resources with:

kubectl apply -f spark-executor-sa.yaml -f pod-exec-role.yaml -f spark-executor-rolebinding.yaml

2. Create the Execution Job #

This Job will run a container with kubectl pre-installed, execute your script in the target Pod, and exit once complete. We use a label selector (-l app=my-notebook) instead of hardcoding the Pod name to handle Pod restarts or deployment rollouts automatically.

Save this as spark-script-job.yaml:

apiVersion: batch/v1
kind: Job
metadata:
  name: spark-script-job
  namespace: spark
spec:
  template:
    spec:
      serviceAccountName: spark-script-executor
      containers:
      - name: kubectl-runner
        image: k8s.gcr.io/kubectl:latest
        command: ["/bin/sh", "-c"]
        args:
        - kubectl exec -n spark -l app=my-notebook --container my-notebook -- python3 myscript.py
      restartPolicy: OnFailure
  backoffLimit: 3  # Retry up to 3 times if the command fails

Apply the Job with:

kubectl apply -f spark-script-job.yaml

3. Monitor the Job #

Check if the Job is running or completed:

kubectl get jobs -n spark

View logs to confirm the script executed successfully:

kubectl logs -n spark $(kubectl get pods -n spark -l job-name=spark-script-job -o name)

Key Notes #

• Resilience: Using the label selector (app=my-notebook) ensures the Job targets any active Pod from your deployment, even if the Pod name changes after a rollout or restart.
• Cleanup: Once the Job completes successfully, you can delete it with kubectl delete job spark-script-job -n spark.
• Customization: Adjust the backoffLimit for more/fewer retries, or modify the command in the Job to pass arguments to your script if needed.

内容的提问来源于stack exchange，提问作者J.C Guzman