Production readiness checklist

Use this as a pre-launch pass. It is the same shape as Railway’s production checklist, mapped to features that exist here.

Performance and reliability

• Region — the project is pinned at create time (default us-east-1). Keep chatty services and their databases in that project.
• Private networking — app-to-database and app-to-app traffic uses service names, not public URLs.
• Replicas — run at least two replicas on public HTTP if you cannot afford a single-instance blip. See horizontal scaling.
• Health checks — set a path so rolling deploys wait for ready. See health checks.
• Database topology — use HA for production data stores. See databases.

Observability

• Open Coroot from the service and confirm logs and metrics.
• Prefer search_logs over a raw tail in MCP.

Quality

• Keep a non-production environment for staging.
• Turn on PR previews if you want a live canvas per pull request.
• Keep topology in stackblaze.yaml if you want git-reviewed infra.
• Know how to roll back a revision.

Security

• Secrets in variables, not in git.
• 2FA on the account.
• IP allowlisting if the dashboard or app should be source-restricted.
• Do not expose databases on the public internet.

Recovery

• Enable backups on stateful databases.
• Know PITR for Postgres if you need point-in-time restore.
• Volumes are per instance — they are not a backup strategy by themselves.