Stuck Deployment rollout alerts
A Deployment rollout replaces old Pods with new ones. If the new Pods cannot
become available, Kubernetes may report ProgressDeadlineExceeded. kwatch
watches that signal so the team can investigate the stalled rollout.
What kwatch detects
The rollout monitor is enabled by default. It reports a Deployment that has exceeded its progress deadline. The alert points to the affected workload; the exact cause still depends on the new Pods and their Events.
| Setting | Default | Purpose |
|---|---|---|
rolloutMonitor.enabled | true | Watch for stuck Deployment rollouts. |
Investigate a stuck rollout
Start with the Deployment and its rollout status:
kubectl rollout status deployment/<name> -n <namespace>
kubectl describe deployment <name> -n <namespace>
kubectl get pods -n <namespace>
Check the Deployment conditions and the new Pods. Image pull failures,
scheduling problems, container crashes, and readiness failures lead to
different fixes. Inspect a failing Pod with kubectl describe pod and its
recent logs before changing the rollout.
Configure or disable the monitor
Use the interactive kwatch manager to change monitor settings. This is the equivalent configuration fragment:
rolloutMonitor:
enabled: true
After changing settings, run kwatch lint and let the manager verify the
workload. See the Kubernetes coverage guide
for related workload signals.