Integration Monitoring

Integration monitoring

When a third party degrades, your service degrades and your dashboards show your own errors. Monitoring integrations separately is what turns an hour of confused investigation into an immediate attribution.

Method

1. Measure each dependency separately. Latency, error rate, and timeout rate per integration, so a problem is attributable at a glance (see observability).
2. Alert on their error rate, not only on yours. A rising provider error rate is your early warning before it becomes user-visible.
3. Track quota consumption. Approaching a limit is a predictable incident that monitoring can prevent (see rate-limit-handling).
4. Watch latency percentiles, not averages. A provider degrading at the tail affects a subset of users invisibly in the mean.
5. Monitor webhook delivery gaps. Absence of events looks identical to quiet, so expected-rate alerts catch silent failures (see webhook-consumption).
6. Subscribe to their status page and changelog. Their incident communication is a signal your monitoring cannot produce.
7. Log correlation identifiers from both sides. Their request id stored with yours is what makes a support conversation productive.

Boundaries

Monitoring detects and does not prevent, so it pairs with resilience (see integration-resilience). You cannot see inside a provider, only their responses to you. Status pages are often updated after customers notice.