Accept a Swarm deployment for production

Prerequisites

Complete installation, external TLS load balancing, production database/backup, monitoring, and on-call ownership. Identify go/no-go and rollback decision-makers.

Prepare the acceptance record

Record cluster/stack, image digest, runtime mode, database/pool class, service replicas/placement, secret version identifiers, public origin, test operator, and rollback window. Exclude secret values.

Run production acceptance

1. Verify three-manager quorum and worker capacity after one planned drain.
2. Confirm immutable images and protected content-hashed secrets.
3. Require direct one-shot migration success and exclusive runtime ownership.
4. Confirm every service converges and role/queue signals advance.
5. Verify private database/pool networking and TLS public proxy/SSE/webhooks.
6. Confirm DNS host = TLS/load-balancer host = NEXTAUTH_URL = normally NEXT_PUBLIC_APP_URL = saved Application URL; no task or internal host appears in redirects or generated links.
7. Confirm /setup was completed through public HTTPS, login and provider callbacks remain on that hostname, and an unrelated host returns 421.
8. Verify database connections, backup, and isolated restore with matching secrets.
9. Run alert, notification, acknowledgement, escalation/assignment where configured, resolution, and status projection.
10. Drain/restart one application worker and confirm recovery within objective.
11. Confirm dashboards/alerts with the operational on-call.
12. Record evidence and explicit go/no-go.

Verify acceptance

Do not accept on service convergence alone. Security, migration, recovery, notification, and end-to-end incident evidence must pass.

Operate it in production

Repeat affected checks after cluster, topology, database, proxy, provider, secret, or release changes. Schedule restore, manager recovery, node drain, capacity, and upgrade drills.

Troubleshooting failed acceptance

Use Swarm troubleshooting, preserve state/logs, correct the underlying layer, and rerun downstream checks. Never bypass deployment validation or weaken TLS/secrets to force a pass.

Change or undo the release decision

Stop new traffic where safe and invoke Upgrade and rollback or data recovery. Application/stack rollback does not reverse schema migration.

Next steps

• Health and metrics
• Application URL and host routing
• Backup and restore