We're preparing to launch a video-based platform with payments and several other features, but we only have about a week left. Our team consists of two backend developers and one frontend developer. I currently maintain separate development and staging servers and am preparing production infrastructure. I provision the environments with Ansible, use a self-hosted Git service and a dedicated 32-core runner, and deploy through CI/CD pipelines. Many of the pipeline configuration files were generated with AI and then adapted by me. What should we review or prioritize before launch? Which infrastructure, security, deployment, and operational mistakes are most important to avoid, especially when there isn't much time left?
3 Answers
Treat the first release as a controlled pilot rather than assuming it is ready for unlimited public traffic. Confirm who owns production support, how incidents are reported, and what happens if the payment provider, storage, database, or video delivery fails. Test rate limits, authorization boundaries, upload handling, payment webhooks, and user-data deletion. Keep an independent backup, verify monitoring from outside the production network, and have a clear way to disable new signups or payments if something goes wrong.
Freeze the feature set and create a launch runbook. Use immutable, versioned build artifacts so production runs exactly what was tested in staging. Make deployments reversible, keep database migrations backward-compatible where possible, and test both rollback and restoring a backup before launch. Store secrets outside the repository, use least-privilege accounts, restrict administrative access, patch the hosts, configure TLS, and make sure logs and alerts exist for failed deployments, authentication problems, payment failures, disk usage, and resource exhaustion.
Even if you keep the current push-based deployment, you can still make it safer with versioned artifacts, approvals for production, limited runner permissions, and a documented rollback procedure.
Bring in an experienced engineer for a focused review immediately. With one week left, don’t try to perfect the entire platform—identify the highest-risk areas first: payment handling, authentication and authorization, secrets, backups, data loss, exposed services, dependency vulnerabilities, and whether you can recover from a failed deployment. Have every pipeline step and Ansible change understood by a human rather than trusting generated configuration blindly.
That makes sense. I’m looking for a practical checklist because I may not know which gaps are most important to look for.
Good point about operational ownership. We should define the failure procedures before launch instead of improvising during an incident.