A deploy strategy that releases new code without any window of unavailability, the table-stakes for any production-facing service.
Zero-downtime deployment is a release strategy where the new version of a service goes live without any user-visible window of unavailability. The standard techniques are rolling updates (replace replicas one at a time behind a load balancer), blue-green (run both versions, flip traffic atomically), and canary (route a small slice to the new version, ramp up gradually). Each is supported natively by Kubernetes, AWS ECS, and most modern orchestrators. The hardest part is database migrations: the schema must be backward-compatible across both versions for the duration of the rollout.
Maintenance-window deploys (take the service down for 10 minutes during a low-traffic window) are no longer acceptable for any service that has real customers in multiple time zones. Zero-downtime deploys are how teams ship daily without burning through the error budget on every release, and how customer-facing services hit 99.9%+ uptime targets in the presence of a healthy deploy cadence.
See the part of the platform that handles zero-downtime deployment in production.