How to Refresh Worker Secrets Without Killing Deployment?

Question

I want to learn how to update secrets in worker pods without killing and recreating the deployment.

Currently pods pull in secrets to as env vars with:

        env:
        - name: SECRET_ACCESS_KEY
          valueFrom:
            secretKeyRef:
              key: secret_access_key
              name: secrets

but this only happens when they startup.

So if there is a need to change a secret I have to:

1. Change the secret in secrets.yaml
2. kubectl apply -f secrets.yaml
3. kubectl delete -f worker-deployment.yaml
4. kubectl apply -f worker-deployment.yaml

I really don't like step 3 and 4 as they terminate jobs in progress.

What is a better workflow for updating env var secrets in place?

Answer 1

There is no way to do a "hot reload" for pod's environment variables.

Although, you do not need to delete and recreate the deployment again to apply the new secret value. You only need to recreate the underlying pods. Some options are:

• kubectl delete pods to recreate them
• Editing some deployment trivial value to trigger a Rolling Update (e.g., changing the terminationGracePeriodSeconds from 30 to 31).
• Use kubectl rollout restart to do a rolling restart on the deployment †

† rollout restart is only available on kubernetes v1.15+

Answer 2

As already mentioned, what you want to do is not possible. However, there is an alternative offered by Kubernetes: mounting ConfigMaps as Volumes. For example

apiVersion: v1
kind: Pod
metadata:
  name: configmap-pod
spec:
  containers:
    - name: test
      image: busybox
      volumeMounts:
        - name: config-vol
          mountPath: /etc/config
  volumes:
    - name: config-vol
      configMap:
        name: log-config
        items:
          - key: log_level
            path: log_level

In this case, the log-config ConfigMap would be mounted as a Volume, and you could access the contents from its log_level entry as the file “/etc/config/log_level” inside the pod.

Changes to the config map are reflected by changes in the files on the Volume, and those changes can, in turn, be watched by your application by using the appropriate functionality in your language.