Web Properties and Real User Monitoring (RUM)
ServiceRadar Real User Monitoring (RUM) pairs a sibling L7 Web Property
inventory (in:websites) with Akamai boomerang.js (boomerang-sr.min.js)
beacon ingestion so you can correlate Core Web Vitals, CDN edge-vs-origin
latency, corporate branch WAN health (prefix_tags), synthetic HTTP/TLS
checks, and backend OpenTelemetry traces in one place.
Why Web Properties are separate from Devices
ServiceRadar's Device inventory (ocsf_devices and DIRE) models L2/L3 hosts
by MAC address, serial number, agent ID, and IP. Public websites are L7
services:
- Behind a CDN such as Akamai or Cloudflare, DNS resolves a site to rotating regional Anycast VIPs.
- Multiple websites frequently share a single Kubernetes Gateway VIP or load balancer IP.
Modeling websites as ServiceRadar.Inventory.WebProperty (in:websites) keeps
rotating CDN VIPs out of ocsf_devices and prevents false DIRE merges, while
still linking each website down to underlying infrastructure via
WebPropertyBinding:
- Kubernetes public endpoints (
HTTPRoute/Ingresshostnames) - Origin devices (operator-declared or direct non-CDN VIP match)
- OpenTelemetry services (via W3C
traceparentcorrelation) - Active synthetic checks (
ServiceCheckand Solution Packs such asio.serviceradar.http-api-and-tls-posture)
Data flow
Public browser beacons are key-gated and scrubbed in Rust, published to NATS JetStream, and written exclusively to the active telemetry store.
All telemetry flows through NATS JetStream first and writes to the single active telemetry store (StarRocks when enabled, CNPG otherwise; never both).
Creating a Web Property
Navigate to Websites (/websites) and click + Add Web Property:
| Field | Description | Example |
|---|---|---|
| Name | Human-readable label for the site. | Example Storefront |
| Canonical Origin | Normalized scheme://host[:port] unique key. | https://shop.example.com |
| Environment | Production, Staging, Development, or Internal. | Production |
| Allowed CORS Origins | Comma- or newline-separated origins permitted to POST beacons. | https://shop.example.com, https://www.shop.example.com |
| Allowed Query Keys (non-PII) | Optional query keys retained after URL scrubbing. All other query parameters and fragments are stripped at ingress. | utm_source, utm_medium, utm_campaign, ref |
| Trace Propagate Origins | Optional cross-origin API origins allowed to receive W3C traceparent headers. Leave blank for same-origin only. | https://api.example.com |
| RUM Sample Rate | Fraction of sessions sampled (0.0 to 1.0). | 1.0 |
| RUM Enabled | Enables or pauses beacon ingestion for this property. | Checked |
| Anonymize IP | Zeroes /24 IPv4 and /48 IPv6 client addresses after prefix_tags and Geo/ASN lookup. | Checked |
After saving, open the property detail page (/websites/:id) to view its
generated rum_site_key (sr_rum_...) and onboarding snippets. You can
rotate rum_site_key at any time; the collector invalidates the previous key
automatically via NATS policy updates.
rum_site_key is a public write-only routing and rate-limiting identifier
embedded in HTML served to browsers. It never grants read access to ServiceRadar
or OTLP ingestion. Internal OTLP tokens are never exposed to browsers.
Privacy and security controls at ingress
The Rust collector enforces these protections before any beacon reaches NATS JetStream:
- Site-key +
Origingate: Unknown or rotated keys and disallowed origins are rejected in memory without querying CNPG. - Payload and rate limits: Request bodies are capped at 64 KiB with
per-key and per-subnet (
/24IPv4,/48IPv6) token-bucket rate limits. - PII scrubbing: Query strings (
?token=...) and fragments (#...) are stripped from page URLs and referrers by default, keeping only explicitly allowlisted non-PII keys. - IP anonymization: Client IPs are matched against
prefix_tagsand Geo/ASN in memory and then truncated (/24//48) before persistence when Anonymize IP is enabled.
Querying RUM data with SRQL and dashboards
ServiceRadar ships a built-in Web RUM dashboard
(cloud.serviceradar.web-rum) showing p75 Core Web Vitals (LCP, INP, CLS,
FCP, TTFB), Edge vs. Origin RTT vs. Browser Render breakdowns, Akamai CDN cache
status, branch prefix_tags / ASN cohorts, and route hotspots.
You can also query Web Properties and beacons directly with SRQL:
in:websites sort:name:asc limit:50
in:rum time:last_1h sort:observed_at:desc limit:50
in:rum time:last_24h cdn_cache_status:MISS lcp_ms:>2500 limit:50