For a zero-downtime upgrade, Rolling Upgrades (A) are the fundamental mechanism; Docker Swarm replaces tasks incrementally (default parallelism is 1), ensuring the service remains partially active. However, the update logic alone isn't enough to prevent outages if the networking layer isn't designed to route traffic away from the specific container being stopped. Hosting the service behind a Virtual IP (VIP) with no session persistence (C) ensures that the Swarm load balancer (Ingress Routing Mesh) can instantly redirect traffic to the remaining healthy replicas. Without session persistence, the load balancer is free to re-route requests immediately, preventing a user from experiencing a timeout if their "sticky" container is being updated.

B. Enable parallel upgrades: Updating all containers simultaneously causes a total service outage while the new images are pulled and the containers are restarted.

D. Update the restart policy: Restart policies manage recovery from unexpected crashes; they do not control the orchestration or sequencing of a planned version upgrade.

E. Ensure replicas set to a minimum of 5: While maintaining capacity is good, the exhibit already shows 5 replicas. This is a scale setting, not a specific upgrade "design approach."

Official Vendor Documentation: Docker Docs, "Deploy services to a swarm," Apply rolling updates to a service. https://docs.docker.com/engine/swarm/swarm-tutorial/rolling-update/

Official Vendor Documentation: Cisco DevNet Learning, 350-901 DEVCOR Study Guide, Chapter 5: "Infrastructure and Automation," Section: "Container Orchestration."

Official Vendor Documentation: Docker Engine CLI Reference, docker service update, Section: "Update strategies and Endpoint Mode."

Academic Resource: "Cloud Native Patterns," Cornelia Davis, Chapter on Rolling Updates and Load Balancing for Resilient Systems.