1. What to monitor? #
Four complementary angles.
Angles #
| Question | Method | Example tool |
|---|---|---|
| Is the site responding, over HTTPS? | Scheduled HTTP/HTTPS synthetic test from several regions | Blackbox Exporter, k6, Uptime Kuma, Datadog Synthetics |
| Is the certificate about to expire? | Reading the TLS expiry date | probe_ssl_earliest_cert_expiry metric |
| Do visitors get a fast site? | Real User Monitoring (LCP, INP, CLS) | Grafana Faro, Datadog RUM, Dynatrace RUM |
| Does the server handle the load? | Metrics and access logs (5xx codes, latency) | Prometheus, Loki, Elasticsearch |
2. HTTPS monitoring with Prometheus #
Open source and simple.
Blackbox Exporter #
# blackbox.yml
modules:
https_2xx:
prober: http
timeout: 10s
http:
valid_status_codes: [200]
fail_if_not_ssl: true
preferred_ip_protocol: ip4# prometheus.yml
scrape_configs:
- job_name: site
metrics_path: /probe
params: { module: [https_2xx] }
static_configs: [ { targets: [ "https://www.example.com", "https://shop.example.com" ] } ]
relabel_configs:
- { source_labels: [__address__], target_label: __param_target }
- { source_labels: [__param_target], target_label: instance }
- { target_label: __address__, replacement: blackbox:9115 }# Alertes
probe_success == 0 # site injoignable
(probe_ssl_earliest_cert_expiry - time()) / 86400 < 14 # certificat expirant sous 14 jours
histogram_quantile(0.95, rate(probe_duration_seconds_bucket[5m])) > 2Good practice: Probe from at least two locations and alert when several fail: a local network outage should not wake up on-call.
3. Going further #
Journeys and real experience.
Journeys and RUM #
- Journey tests (login, cart) with Grafana k6 in browser mode, scheduled or in CI.
- RUM with Grafana Faro for real Core Web Vitals per page, browser and country.
- A public status page fed by the same probes.
- Correlation: a failing probe should lead to the backend's logs and traces.
