Telemetry
Dashboard example

API reliability and latency dashboard

Compare traffic, error rate, and tail latency by endpoint while retaining enough volume to judge whether a percentile is stable.

Reviewed by the Telemetry product team on . Decision, event grain, metrics, DataFusion SQL, synthetic result, assumptions, freshness, and interpretation boundary. Review standards and ownership

Row grain

One terminal request outcome before grouping by route template.

Decision supported

Prioritize an endpoint for release comparison, dependency inspection, or rollback review.

Metric contract

Put the denominator and units beside the chart

Adapt these names to your contract while preserving the declared grain. Inspect the linked event schema before mapping fields from production producers.

  • Request volume
  • Error rate
  • p50 latency
  • p95 latency
Inspect the event schema
Complete DataFusion SQL

Review the query before adapting the fields

This public query is a starting definition. Add a bounded time filter, complete-bucket policy, environment, and minimum volume appropriate to your production event contract.

SELECT
  endpoint,
  COUNT(*) AS requests,
  ROUND(
    100.0 * SUM(CASE WHEN status = 'error' THEN 1 ELSE 0 END)
    / NULLIF(COUNT(*), 0),
    2
  ) AS error_rate_pct,
  approx_percentile_cont(latency_ms, 0.50) AS p50_ms,
  approx_percentile_cont(latency_ms, 0.95) AS p95_ms
FROM api_requests
GROUP BY endpoint
ORDER BY error_rate_pct DESC, requests DESC;
Data assumptions

Prove these before publishing the result

  • Endpoint values are bounded route templates rather than raw URLs.
  • Latency uses one end-to-end millisecond definition across producers.
  • Operational charts omit the newest incomplete time bucket.
Query review

Checks that keep the dashboard trustworthy

  • Add a bounded time filter before using this query in production.
  • Keep request volume beside every percentile and rate.
  • Segment by release and error type after detecting a regression.

Interpretation boundary

What this result cannot prove by itself

Always show request count beside a percentile. A high p95 from a tiny group is not equivalent to broad customer impact.

Related dashboard examples

Validate the definition with known data

Run a fixture with known success, failure, missing, duplicate, and boundary cases before connecting the query to a production dashboard or alert.