---
name: uptime-and-monitoring-setup
description: Use when asked how the owner will know a website or web app is down, broken, slow, or expiring - uptime and keyword checks, form and key-flow checks, TLS certificate and domain expiry, error tracking, Core Web Vitals field monitoring, build/deploy failure alerts, and who gets alerted. Designs a lean monitoring plan using free tiers where possible and proposes config as a review PR. Does not sign up for services or spend money without approval.
---

# Uptime and monitoring setup

Last reviewed: 2026-11-10

## Quick prompts

For anyone improving a site or app, whether a small-business owner or someone at a larger company practicing better prompts. Without monitoring, the first person to notice the site is down is often a frustrated customer. Paste a prompt into any AI assistant that can see the site or repo.

1. "For the site I work on, list what 'broken' would mean (home page down, form not sending, and so on) and suggest one simple check for each. Plan only; don't sign up for anything."
   *Why this prompt works:* Defining 'broken' first gives the assistant clear criteria.

2. "Suggest a free or low-cost way to get alerted when the site I work on goes down or its security certificate or domain is about to expire. Explain the options plainly before setting anything up."
   *Why this prompt works:* Setting a budget limit keeps suggestions realistic.

3. "Who should get alerts for the site I work on, and how, so they aren't ignored? Draft a short alert plan with a main and backup contact for me to review."
   *Why this prompt works:* Asking for a draft keeps you in charge of the final decision.

## Hard limits

- Do not create accounts, enter payment details, or add the owner's phone/email to a service without explicit approval.
- Synthetic checks must not submit real forms repeatedly or create junk records; use a test mode, a health endpoint, or a dedicated test inbox the owner approved.
- Keep alerts few and actionable; an ignored alert is no alert. Never put secrets in monitor URLs.

## Inputs to gather

1. What "broken" means for this site: home down, a key page down, search broken, forms not delivering, checkout failing, login failing.
2. Who should be alerted, how (email, chat, SMS), and quiet hours.
3. Hosting/CDN features already available (analytics, health checks, notifications on failed builds).
4. Domain registrar and renewal settings; certificate issuer (usually automatic).

## Interface-type notes

- **Content sites**: uptime on home plus one article; keyword check that the page body contains expected text (catches blank or error pages that return 200); scheduled-build success; sitemap reachable.
- **Marketing**: landing pages and form delivery; tag manager failures do not take the page down but break attribution.
- **E-commerce**: checkout synthetic in test mode, payment provider status page, inventory sync jobs.
- **SaaS / app**: login synthetic with a test account, API health endpoint, background job queues, error rate from error tracking.
- **SPA**: a 200 shell can hide a broken app; check for a rendered element with a browser-based check.

## Steps

1. **Map critical paths** and pick one check for each (HTTP status, keyword, browser step, or heartbeat for scheduled jobs).
2. **Choose tools**: host's built-in notifications, a free uptime monitor, GitHub Actions scheduled workflow for simple checks, error tracking with a free tier, real-user Web Vitals via the `web-vitals` library into existing analytics.
3. **Expiry**: TLS certificate expiry alert (even with auto-renewal), domain expiry and auto-renew on, payment card on the registrar valid (owner checks).
4. **Scheduled jobs**: heartbeat ("dead man's switch") monitors for cron builds and backups so silence triggers an alert.
5. **Alert routing**: one primary contact, one backup, a short note in each alert saying what to check first (link to the runbook).
6. **Status and runbook**: optional public status page; a one-page "site is down" runbook linking to backup-and-rollback-readiness.

## Validation

- Trigger each alert once in a safe way (monitor a test URL that returns 500, pause a heartbeat) and confirm delivery to the right person.
- Confirm checks do not count toward rate limits or analytics in a misleading way (exclude monitor user agents from analytics if possible).

## Report format

1. Critical paths and the check for each (type, URL/endpoint, interval, expected result).
2. Tools proposed with cost (free tier limits) and what needs owner sign-up.
3. Alert routing table. 4. Config as a review PR (for example a scheduled workflow), "not merged". 5. Test results of each alert.

## References

- Google SRE book, Monitoring Distributed Systems: https://sre.google/sre-book/monitoring-distributed-systems/
- web-vitals library: https://github.com/GoogleChrome/web-vitals
- GitHub Actions scheduled workflows: https://docs.github.com/en/actions/writing-workflows/choosing-when-your-workflow-runs/events-that-trigger-workflows#schedule
- Let's Encrypt expiration emails ending, so monitor yourself: https://letsencrypt.org/2025/01/22/ending-expiration-emails/
- ICANN domain renewal basics: https://www.icann.org/resources/pages/domain-name-renewal-expiration-faqs-2018-12-07-en
