Rolling back deployments

When things go wrong in production after a deployment, the best-practice response is to roll back the changes and get back to the last previous version known to work. The way you go about this depends on the deployment pattern you've employed. Let's consider them one by one.

Rolling back standard Kubernetes deployments

Kubernetes deployments keep a history. For example, if we edit the user manager deployment and set the image version to 0.5, then we can see that there are two revisions now:

$ kubectl get po -l svc=user,app=manager -o jsonpath="{.items[0].spec.containers[0].image}"
 g1g1/delinkcious-user:0.5
 
 $ kubectl rollout history deployment user-manager
 deployment.extensions...