Minimizing interruptions ensures that clients and applications continue to function without noticeable delays. The following table describes common patterns that help reduce downtime during upgrades or deployments. Each pattern includes a description and its trade-offs:
Pattern Description Trade-off
Side-by-side compose Start a new stack on an alternate port and then switch the client to the new stack Requires dynamic endpoint switching or a load balancer
Blue or green DNS Maintain two DNS targets and flip traffic after health checks pass Requires tuning DNS TTL values for fast propagation
In-place rolling Stop the old stack and start the new stack quickly Causes brief unavailability unless a load balancer manages multiple instances