Guide & TutorialsMonitor your API integration
Guide & Tutorials
Monitor your API integration
Turn request metadata into useful dashboards and actionable alerts.
Production integrations need more than error logs. Good telemetry shows whether customers are affected, whether failures are temporary, and where an investigation should begin.
Photo by Lukas Blazek on Unsplash.
Capture useful context
Record the endpoint template, HTTP method, response status, duration, retry count, and Northstar request identifier for every API call. Use endpoint templates such as
/projects/{id} rather than raw URLs so resource identifiers do not create excessive metric cardinality.json
{
"service": "northstar-client",
"method": "POST",
"endpoint": "/projects/{id}/deployments",
"status": 201,
"duration_ms": 284,
"request_id": "req_01J8ZA8E1T"
}Never include bearer tokens, webhook secrets, or complete customer payloads in logs.
Measure service health
Track request volume, error rate, and latency together. A low error rate can still hide a serious problem if latency rises enough to trigger timeouts in your application.
| Signal | What it reveals |
|---|---|
| Request rate | Traffic changes and unexpected integration loops |
| Error rate | Failed requests grouped by endpoint and code |
| Latency | Slow dependencies and timeout risk |
| Retry rate | Transient instability before requests fail permanently |
| Rate-limit capacity | How close the integration is to throttling |
Design actionable alerts
Alert on sustained customer impact rather than isolated failures. A single
500 may recover automatically; a rising error ratio over several minutes deserves attention.Warning
Do not retry every failure. Validation and authorization errors require a configuration or code change and will not recover through backoff.
Investigate efficiently
Begin with the time window and affected endpoint, then inspect error codes and request identifiers. The request identifier lets Northstar support trace the server-side request without exposing sensitive application data.
