Skip to content
CCinder
Understanding Fallback Telemetry

Guides

Understanding Fallback Telemetry

Inspect fallback reasons, timeout/error classification, and usable vs unusable normalized results.

Fallback should be observable, not mysterious.

Subscribe to fallback events

ts
7 lines
Cinder.on('fallback:start', (payload) => {
  console.log(payload.reason, payload.provider, payload.timeoutMs);
});

Cinder.on('fallback:end', (payload) => {
  console.log(payload.errorCode, payload.used, payload.latencyMs);
});

What to inspect

fallback:start

Look at:

  • reason
  • provider
  • authMode
  • timeoutMs

fallback:end

Look at:

  • used
  • normalized
  • errorCode
  • statusCode
  • latencyMs
  • errorMessage

Why this matters

Telemetry helps you tell the difference between:

  • local unknowns
  • provider/network failures
  • timeout failures
  • invalid provider responses
  • normalization failures against the real graph

Continue reading