Deployment
Validate API latency, uptime, error spikes, and launch-window deployment health.
Integration
Deployment
API latency, uptime, error spikes, and launch window monitoring.
Where the Environment check asks "is the surface configured correctly?", the Deployment check asks "is the running app actually healthy under real conditions?" PreFlight samples API latency, builds uptime percentages over time, records error spikes, and watches the launch window so post-deploy regressions surface before customers feel them.
Requirements
- Project with a saved health endpoint or production URL
- App live and responding at that URL
- Optional Monitoring or Sentinel enabled for repeated sampling—not just a single snapshot
What PreFlight checks
| Probe | What it proves |
|---|---|
| API latency | The health endpoint responds within an acceptable time budget. |
| Uptime sampling | Repeated samples via Monitoring or Sentinel build a real uptime percentage. |
| Error-rate signal | Non-2xx responses and timeouts are recorded and can trigger alerts after repeated failures. |
| Launch window | During 24-hour Monitoring, samples run on a cadence to catch post-deploy regressions. |
| Cold-start signal | Serverless cold starts and region latency show up in response-time trends. |
Failure guidance
| Symptom | Fix |
|---|---|
| Latency over budget | Profile the slow route, check cold-start behavior, and align database/cache region with deployment. |
| Error spike after deploy | Open Incidents and the Flight Recorder to correlate the spike with the exact deploy, then roll back or patch. |
| No samples recorded | Confirm Monitoring or Sentinel is enabled and the project has a reachable health URL. |
| Intermittent timeouts | Check upstream dependencies (database, Redis, third-party APIs) and increase health check timeout if appropriate. |
| Uptime below target | Review recent probe history, fix recurring failures, and enable Sentinel for continuous coverage. |
How it relates to the watchdogs
Monitoring runs deployment health samples on a cadence during a 24-hour release window—start it right after a risky deploy. Sentinel continues the same sampling indefinitely once the launch window closes. Vercel Watch adds deployment-platform awareness: which build is live and whether the production alias drifted.
Dashboard setup
- Save a health endpoint or production URL in Project Settings.
- Run an initial Pre-Flight Check to establish a latency baseline.
- Start 24-hour Monitoring immediately after a risky deploy.
- Enable Sentinel when the launch window closes for always-on uptime sampling.