A development team wants to deploy a new version with minimal downtime and quick rollback. Which deployment approach should be used?

A development team wants to deploy a new version with minimal downtime and quick rollback. Which deployment approach should be used?

• A. Use a rolling deployment to update all the servers in the PROD cloud environment with the new application. To switch to the previous version, repeat the process.
• B. Deploy the application to PROD and the previous version to QA. To switch back, promote the QA environment to PROD.
• C. Deploy the application to a subset of servers and route traffic to these servers. To switch to the previous version, change the route to the non-updated servers.
• D. Deploy the application to a staging environment and force a failover to this environment. To restore the previous version, restore from the previous night’s backup.

Answer: (A)

Rolling deployments minimize downtime by updating servers in small batches while the rest of the system keeps serving users. As each batch is updated, the service remains available, and if a problem emerges, you can roll back by redeploying the previous version across the updated servers or by reversing the rollout to restore the old version. This provides a straightforward, proven path to quick rollback without needing a separate production clone, which is especially important when you want to keep the production environment online throughout the process. Other approaches introduce extra complexity or risk. Deploying to QA and then promoting that environment to production can create data and configuration drift and isn’t as clean a rollback path. A canary-like approach that routes traffic to a subset of servers can also work and reduce risk, but it relies on precise traffic shifting and monitoring, which can complicate back-out in practice. The rolling deployment approach directly satisfies both minimal downtime and an easy rollback in a straightforward way.

Rolling deployments minimize downtime by updating servers in small batches while the rest of the system keeps serving users. As each batch is updated, the service remains available, and if a problem emerges, you can roll back by redeploying the previous version across the updated servers or by reversing the rollout to restore the old version. This provides a straightforward, proven path to quick rollback without needing a separate production clone, which is especially important when you want to keep the production environment online throughout the process.

Other approaches introduce extra complexity or risk. Deploying to QA and then promoting that environment to production can create data and configuration drift and isn’t as clean a rollback path. A canary-like approach that routes traffic to a subset of servers can also work and reduce risk, but it relies on precise traffic shifting and monitoring, which can complicate back-out in practice. The rolling deployment approach directly satisfies both minimal downtime and an easy rollback in a straightforward way.