Admin Dashboard — Observability

Status: populated Owner: Platform Engineering (Frontend) Last updated: 2026-04-18

1. Logging

Structured JSON via pino in the Next.js server runtime.

Log Format

{
  "time": "2026-04-18T12:00:00.000Z",
  "level": "info",
  "service": "admin-dashboard",
  "requestId": "req_01H...",
  "adminUserId": "usr_01H...",
  "method": "DELETE",
  "path": "/api/operators/op_01H...",
  "status": 204,
  "durationMs": 87
}

Admin user ID is included in all log lines for audit traceability. JWTs, passwords, and SMPP credentials are never logged.

2. Metrics

GET /api/health for Kubernetes probes.

Metric	Type	Labels
admin_request_total	Counter	method, path, status
admin_request_duration_ms	Histogram	method, path
admin_poll_total	Counter	target, result
admin_upstream_error_total	Counter	upstream, status
admin_auth_refresh_total	Counter	result

3. Error Tracking

Sentry with:

• beforeSend hook strips Authorization headers, cookie values, SMPP passwords.
• Admin user ID attached to Sentry scope for all events (non-PII for internal users).

4. Dashboard Polling Observability

The admin_poll_total counter tracks each 30-second poll cycle:

• result=success: metrics data returned and rendered.
• result=stale: upstream returned 304 or data unchanged.
• result=error: upstream returned non-2xx; stale data left in place.

5. Alerting

Alert	Condition	Severity
AdminDashboardHighErrorRate	admin_upstream_error_total{status=~"5.."} > 3/min for 3 min	warning
AdminDashboardPollFailureSpike	admin_poll_total{result="error"} > 5 consecutive	warning
AdminDashboardAuthFailure	admin_auth_refresh_total{result="failure"} > 3/min	critical
AdminDashboardPodNotReady	Readiness probe failing > 2 min	critical

6. Distributed Tracing

OpenTelemetry traces emitted from server component fetch calls. X-Admin-User-Id propagated to Kong and upstream services for end-to-end trace correlation. Admin traces are stored with higher retention (30 days vs. 7 days for customer traces) for audit purposes.