Monitoring & alerts
Keep an eye on a running pipeline without babysitting it manually.
Connector health
Poll a connector's health field via the Connectors endpoint, or better, subscribe to connector.error so you're notified the moment it degrades.
Subscribing to pipeline lifecycle events
The most reliable way to monitor a pipeline is a webhook subscription on pipeline.failed and connector.error.
subscribe.sh
bash
curl -X POST https://api.kmp.kaiju.go.example/v2/webhooks \
-H "Authorization: Bearer $KMP_API_KEY" \
-d '{"url":"https://you.example.com/hooks/kmp","events":["pipeline.failed","connector.error"]}'Recommended baseline
At minimum, subscribe to pipeline.failed for every production pipeline. It's the single highest-signal event for catching problems before customers do.
Dead-letter queue depth
A growing dead-letter queue is an early warning sign, not just a symptom — see Handling errors & retries for how events end up there and how to drain them.