Rollout Health

Rollout Health

Readiness Gates and Rollout Health

A rollout advances only on Ready Pods, so readiness is its brake. Point the probe at a path the API lacks, and the first new Pod never turns Ready, no old Pod is removed, and after progressDeadlineSeconds the rollout fails:

A rollout stopped by readiness and its progress deadlineShell

kubectl patch deployment api --type=json -p='[{"op": "replace", "value": "/readyz",
  "path": "/spec/template/spec/containers/0/readinessProbe/httpGet/path"}]'
kubectl rollout status deployment/api --timeout=120s
kubectl get pods -l tier=api
kubectl get deployment api -o jsonpath='{.status.conditions[?(@.type=="Progressing")].reason}'
echo && kubectl rollout undo deployment/api

Output

deployment.apps/api patched
Waiting for deployment "api" rollout to finish: 1 out of 3 new replicas have been updated...
error: deployment "api" exceeded its progress deadline
NAME                   READY   STATUS    RESTARTS   AGE
api-54b58bdffd-5sk4n   0/1     Running   0          61s
api-6d799c4498-dvdgn   1/1     Running   0          81s
api-6d799c4498-pllhc   1/1     Running   0          87s
api-6d799c4498-v25v7   1/1     Running   0          71s
ProgressDeadlineExceeded
Warning: resource deployments/api was previously managed with 'kubectl apply'. Rolling back
  will not
  update the kubectl.kubernetes.io/last-applied-configuration annotation, ...
deployment.apps/api rolled back

The old Pods served throughout. CI should fail on ProgressDeadlineExceeded (default deadline 600 seconds); nothing rolls back by itself. A readiness gate (spec.readinessGates) adds Pod conditions an outside controller must set, such as the AWS 24 Load Balancer Controller's health check.

MENU