You update a Deployment, run kubectl rollout status, and wait. The command times out. Yet when you send a request to the Service, it still answers. Did the update finish? If not, which Pods are serving your traffic?

In this disposable Kubernetes lab, you'll investigate that exact scenario. Starting from a healthy application, you'll introduce three failures: an image that can't be pulled, a readiness probe that never passes, and a Pod that can't be scheduled. You'll identify the blocked revision, inspect its evidence, and verify recovery before moving on.

In the recorded image failure, the Deployment reported both of these conditions. This table is derived from its JSON snapshot:

Condition Status Reason
Available True MinimumReplicasAvailable
Progressing False ProgressDeadlineExceeded

Those conditions answer different questions. The following experiments show why an application can still answer requests while its update is stuck.

Table of Contents