Overview
In a rolling update, the deployment system (like Kubernetes) updates instances one by one (or in small batches). It waits for the new instances to become healthy before terminating the old ones. This ensures that there is always enough capacity to handle incoming traffic.
Key Features
- Zero Downtime: The application remains available throughout the update.
- Automatic Rollback: Many systems can automatically roll back if the new instances fail health checks.
- Configurable Pace: Users can control how many instances are updated at a time (max surge and max unavailable).
Comparison
Unlike Blue-Green, it doesn't require doubling the infrastructure, but it means that two different versions of the application will be running simultaneously for a period.