Rolling Update

Replace old instances with new instances gradually.

deploy

Detailed Description

Rolling updates are simple and common in Kubernetes.

APIs, events, and database schemas must be compatible across old and new versions during rollout.

Tune maxSurge and maxUnavailable to balance speed versus capacity and availability.

Readiness probes gate traffic to only healthy new pods; use liveness probes to restart unhealthy ones.

Rollback is usually kubectl rollout undo, but recovery can take minutes on large replica sets.

[v1][v1][v1]
-> [v2][v1][v1]
-> [v2][v2][v1]
-> [v2][v2][v2]

Pros

No duplicate full environment, simple default

Cons

Old and new versions coexist during rollout

Examples: Kubernetes Deployment default rollout