DevOps · interactive
Kubernetes rolling deploys
A rolling update swaps pods one at a time, and only sends traffic to a new pod once its readiness probe passes. Roll out a good version, then a broken one, and watch the old pods keep serving.
Deployment “web” · replicas 4
progressing
Service traffic
Only Ready pods get requests. 4 of 4 pods are Ready.
v1-1v1
Ready
v1-2v1
Ready
v1-3v1
Ready
v1-4v1
Ready
Tick 0: 4 pods of v1 are Ready and serving.
New version
maxSurge
1 maxSurge lets the Deployment run up to 5 pods for a moment, so it never has to stop an old one first.
0 v2 pods created so far.
2 3 4 5
In production
Tools and technologies
Rolling updates, readiness probes and rollbacks are the same idea on every orchestrator and cloud.
Not tied to one language — use these names when you search docs or design a system.
- Kubernetes DeploymentsPlatform
maxSurge, maxUnavailable, readiness probes, rollout undo.
GitHub ↗ - Argo RolloutsSoftware
Canary and blue-green with automatic rollback on bad metrics.
GitHub ↗ - FlaggerSoftware
Progressive delivery that promotes or rolls back from metrics.
GitHub ↗ - Amazon ECSService
Rolling deployments with health checks and circuit breaker rollback.
Official site ↗