On kubernetes helm how to replace a pod with new config values

Viewed 23538

I am using helm charts to deploy pods with a "ConfigMap" managing the configurations.

I edit ConfigMap directly to make changes to configuration files and then delete pods using kubectl delete, for the new configuration to take effect.

Is there any easy way using helm to replace a running pod with the new configuration without executing "kubectl delete" command

4 Answers

We have found that using --recreate-pods will immediately terminate all running pods of that deployment, meaning some downtime for your service. In other words, there will be no rolling update of your pods.

The issue to address this in Helm is still open: https://github.com/kubernetes/helm/issues/1702

Instead helm suggests adding a checksum of your configuration files to the deployment in an annotation. That way the deployment will have a different hash and essentially look 'new' to helm, causing it to update correctly.

The sha256sum function can be used to ensure a deployment's annotation section is updated if another file changes:

kind: Deployment
spec:
  template:
    metadata:
      annotations:
        checksum/config: {{ include (print $.Template.BasePath "/configmap.yaml") . | sha256sum }}
[...]

From the docs here: https://helm.sh/docs/charts_tips_and_tricks/#automatically-roll-deployments-when-configmaps-or-secrets-change

@Oliver solution didn't work for me because pods were not recreated by updating the deployment annotations.

The solution is to use dynamic config map names based on your values.yaml file.

In values.yaml:

configVersion: # Change those numbers to force recreating pods
  myApp: 1

In your config map:

metadata:
  name: {{ .Release.Name }}-my-config-v{{ .Values.configVersion.myApp}}

In your deployment:

- configMapRef:
  name: {{ .Release.Name }}-my-config-v{{ .Values.configVersion.myApp}}
Related