Before launching a website, configure 15 concrete checks across the public HTTP path, critical pages, content, API, certificates, DNS, domain registration, ports and backup heartbeat. Define a pass condition for each and trigger a safe failure in staging. A green homepage alone does not cover the user path.

This checklist assumes a site with an API, email and scheduled backups. Replace irrelevant items with your own critical dependency. enterno’s MONITOR_TYPES registry currently includes HTTP, SSL, ping, DNS, domain, heartbeat, port, API, dns_change, UDP, visual and RUM. Some types appear more than once below on purpose: separate homepage and checkout HTTP monitors produce distinct alerts. Create them in the monitor dashboard; plan and region availability vary.

Check the user path before dependencies
Start with what a visitor opens. Google SRE describes black-box monitoring as an outside symptom check. Internal CPU and logs help locate causes but do not replace the user result. HTTP status also needs content: RFC 9110 does not know whether your product card or sign-in button works.
- Homepage. External HTTP GET with expected final 200. Pass when the response is stable from the primary probe and redirects land on the canonical host.
- Sign-in page. Separate HTTP monitor without user cookies. Pass when the form loads without 500 or redirect loops. This is not an actual login test.
- Critical catalog or product. Monitor one public path that matters to users. Pass when its content is available; the homepage cannot stand in for it.
- Checkout or lead path. GET the safe starting page without creating an order. Pass when the form loads. Test the transaction separately in a controlled synthetic flow.
- Static asset. HTTP check on essential JS/CSS or a public manifest. Pass when the file returns its expected status and is not an HTML placeholder.
- Required content marker. HTTP keyword
containson a critical page. Pass when a stable server-rendered marker appears. See the HTTP 200 content guide. - Forbidden error marker. Separate HTTP keyword
not_contains. Pass when a staging error page correctly marks the monitor down. - API health. API monitor with expected status and
assert_bodywhen the contract has a safe marker. Pass without a real user’s secret token. - Primary TLS certificate. SSL monitor on the public hostname. Pass when name and chain are correct and expiry leaves renewal time. Let’s Encrypt describes certificate-monitoring options.
- API or second-host TLS. Separate monitor if another CDN or load balancer terminates TLS. Pass when the certificate users see is verified, as in the s_client guide.
- Site address DNS. DNS monitor for the public A/AAAA answer. Pass when address and delegation match the rollout plan. Cross-check with dig.
- Mail or NS records. Separate DNS change check for MX/NS when email matters. Pass when a test change creates the expected signal and routine changes do not leave false incidents.
- Domain expiration. Domain monitor. Pass when renewal date and contact are known; EPP statuses distinguish a registry hold from a DNS error.
- Critical TCP port. Port monitor for a service meant to be public. Pass when a reachable port is not mistaken for a functioning application; test the protocol separately.
- Backup job. Heartbeat monitor for the scheduled job. Pass when a ping follows a checked archive and a safely missed staging run alerts. See the backup script.
Do not expose a private port merely to tick a checklist box. Replace nonexistent API, email or payment items with your own critical path. For a multi-country audience, add independent regions to the homepage monitor. The regional alert guide explains why disagreement is a partial outage rather than global failure.
Test alert delivery, not only green cards
| Signal | Safe test | Pass condition |
|---|---|---|
| HTTP/content | Replace a staging response | Down and recovery appear in history |
| DNS/TLS | Test domain or certificate, never the sole production one | Alert names the actual cause |
| Heartbeat | Skip a staging job run | Alert arrives after configured grace |
| Notification channel | Supported test message | On-call person receives it in the intended channel |
Verify recipients, quiet hours and incident ownership. Fifteen green monitors without a working notification destination are not launch-ready. Do not let the only production certificate expire or skip the only backup to test an alert. For critical user paths, add a latency threshold after measuring a normal distribution using curl timing, not an invented number.
A live prelaunch spot check
On November 14, 2026, a safe external check of enterno.io returned HTTP 200 in 0.273 seconds from our node, A address 81.163.20.249, TLS 1.3 and Verify return code: 0 (ok). These separate layer measurements do not prove sign-in or backup health and are not an SLA report. Commands and UTC are in data.json.
curl: http=200 total=0.272767
DNS A: 81.163.20.249
OpenSSL: Protocol: TLSv1.3
OpenSSL: Verify return code: 0 (ok)
How we made this guide
We mapped 15 items to src/monitor-types.php and the HTTP/API/heartbeat settings rules. Three public commands ran on November 14, 2026 without customer data. This is a sample verification, not a claim that all 15 monitors were created for enterno.io. SerpAPI snapshots cover Google US and Yandex Moscow. Wordstat region 225 returned 2 for the exact Russian phrase and 38 for a broader checklist query.
FAQ
Does every site need all 15 monitors?
No. Replace absent API, email or payment paths with your real dependencies. Keep a testable pass condition for each item.
Why is the port open while the site is broken?
A TCP connection confirms only port reachability. Check HTTP and content separately.
Can GET test a real login?
No. It checks the form only. A login needs a safe synthetic flow and dedicated test account.
When is the launch monitoring ready?
When critical checks pass, staged failures reach on call and recovery closes incidents correctly.