Files
AndreaSlopFotografie/aidlc-docs/features/frontend-shell/operations/monitoring/monitoring-setup.md
T
SluijsensandClaude Opus 5.5 6d3a1566c2
Continuous Integration / lighthouse (pull_request) Canceled after 0s
Continuous Integration / deploy-test (pull_request) Canceled after 0s
Continuous Integration / deploy-production (pull_request) Canceled after 0s
Continuous Integration / check (pull_request) Canceled after 2m22s
Adds Sentry and Umami monitoring to the site
Sentry loads after the page and reports errors only, without personal data;
Umami counts visits per environment. Both only run on andreaslop.nl and
test.andreaslop.nl. The CSP allows just these two hosts, and builds carry the
commit SHA as release.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 12:08:47 +02:00

5.6 KiB

Monitoring Setup — frontend-shell

What to do once in each service, and what changes in the code. The code changes are made on PR #4 after approval of this setup. Until the values below are filled in, the site sends nothing (safe default).

1. Sentry (logging + error e-mails)

In Sentry (once)

  1. In your existing organisation, create a project: platform React, name andreaslop-fotografie.
  2. Copy the DSN (Settings → Client Keys). It looks like https://<key>@o4511795618185216.ingest.de.sentry.io/<project-id>.
  3. Settings → Security & Privacy: turn on "Prevent Storing of IP Addresses" (M5).
  4. Alerts: keep the default rule "A new issue is created" → e-mail to you (Q7 = A). Environments test and production appear automatically after the first event.

In the code (after approval)

File Change
package.json Add @sentry/react (same package as SlpSoftware)
src/monitoring/config.ts (new) sentryDsn (from step 2) and monitoringFor(hostname) → { environment, umamiWebsiteId } or null for unknown hosts (M1)
src/monitoring/sentry.ts (new) startSentry(): loads @sentry/react dynamically after the load event (M4); init with environment, release (commit SHA from the build), all dataCollection categories off (SDK 11 replaced sendDefaultPii and collects everything by default), no tracing or replay. reportError(error): queues errors, including unhandled ones, until Sentry is ready.
src/monitoring/sentryClient.ts (new) Re-exports only init and captureException, so the lazy chunk is tree-shaken (31 kB gzip instead of 150 kB).
src/main.tsx Calls startSentry()
src/components/StatusPage/ErrorPage.tsx Besides console.error, also reportError(error) (BR-ERR-03 extended)
config/securityHeaders.ts connect-src + Sentry ingest host (M7); regenerated into web.config and nginx
Workflow check job Build with VITE_RELEASE=${{ github.sha }} so Sentry shows which release an error came from

Check after deploy

On https://test.andreaslop.nl, run in the browser console:

setTimeout(() => { throw new Error('Sentry-test vanaf test') })

The error appears in Sentry under environment test within a minute, and you get an e-mail for the new issue.

2. Umami (visitor dashboard)

In Umami (once)

  1. Log in at https://analytics.slpsoftware.nl.
  2. Settings → Websites → Add website, twice:
    • name Andrea Slop Fotografie, domain andreaslop.nl
    • name Andrea Slop Fotografie (test), domain test.andreaslop.nl
  3. Copy both Website IDs (a UUID per website).

In the code (after approval)

File Change
src/monitoring/config.ts The two website IDs, one per hostname (M6)
src/components/Analytics/UmamiAnalytics.tsx (new) Adds https://analytics.slpsoftware.nl/script.js with data-website-id, data-domains=<hostname> and data-do-not-track="true", only for a known hostname (M1). Same pattern as SlpSoftware's UmamiAnalytics.
src/components/AppShell.tsx Renders UmamiAnalytics
config/securityHeaders.ts script-src and connect-src + https://analytics.slpsoftware.nl (M7)

Check after deploy

Open https://test.andreaslop.nl. In Umami, the test website shows a visitor under "Realtime". If you have Do Not Track switched on in your browser, you won't see yourself; that is expected.

3. UptimeRobot (availability + down e-mails)

No code needed. Create these monitors in your existing account (type HTTP(s), interval 5 minutes, alert contact: your e-mail):

Monitor URL When
Andrea Slop — test https://test.andreaslop.nl/ Now
Andrea Slop — test — deep link https://test.andreaslop.nl/portfolio Now (checks the SPA fallback)
Andrea Slop — productie https://andreaslop.nl/ Now. It watches the current WordPress site until the DNS switch, and the new site after that.
Andrea Slop — productie — deep link https://andreaslop.nl/portfolio After go-live. On WordPress this URL may not exist and would give false alarms.

Optional: type Keyword with the keyword Andrea Slop Fotografie. That catches an empty or wrong page that still answers 200; the word is in the static HTML of the new site.

4. Tests (part of the code changes)

  • monitoringFor: example tests for each known hostname, plus a property test that every other hostname (random strings, localhost, IP addresses) returns null (PBT-03 invariant).
  • UmamiAnalytics: no script for an unknown host; the right data-website-id and data-domains for each known host; only one script element.
  • reportError: errors before initialisation are queued and sent once Sentry is ready; nothing is sent for an unknown host.
  • securityHeaders: the production CSP contains exactly these two external hosts, in the right directives only, and no wildcards.
  • serverConfig: drift test; web.config and nginx contain the new CSP.

5. Values (filled in 2026-10-19, in src/monitoring/config.ts)

Value Setting
Sentry DSN https://12ec77dcc3216a98b96d3fb3dac13f66@o4511795618185216.ingest.de.sentry.io/4512282287472720 (project andreaslop-fotografie)
Umami website ID, production 3edbb7d1-4388-4f2d-9cef-11ade88ace96
Umami website ID, test 7fb301ab-6ecc-4aea-beee-71bcb6009bfe

Sentry project created via the Sentry MCP. The default alert ("high priority issues") was changed to "a new issue is created" → e-mail, to match Q7 = A: handled errors from reportError often get a lower priority and would not have sent an e-mail.