What is error rate, and how do you measure it in Metabase?

Error rate is the share of failed events — errors over total requests, sessions, or users, per window. It's the metric that turns raw error volume into something comparable across services and releases. Measure it in Metabase from rollups synced from Sentry, Prometheus, Datadog, or Cloudflare.

What error rate measures

It measures failure density, normalized for traffic. Raw error counts rise with growth even when nothing got worse; error rate separates "more users" from "more broken." Trended by service it feeds availability and SLO reporting; segmented by release it becomes the core of release-quality analysis.

What data does it need?

• A rollup table with both numerator and denominator per window: error_count(or total - successful) and total_requests / sessions.
• service, environment, and optionally release_id for segmentation.
• Source: Prometheus query_range exports, Sentry event rollups plus session counts, Datadog timeseries, or edge analytics like Cloudflare's.

SQL patterns

SELECT
  service_name,
  date_trunc('week', window_start) AS week,
  ROUND(
    100.0 * SUM(total_requests - successful_requests)
    / NULLIF(SUM(total_requests), 0), 3
  ) AS error_rate_pct
FROM service_health_rollups
WHERE environment = 'production'
GROUP BY 1, 2
ORDER BY 1, 2;

SELECT
  r.version AS release,
  SUM(e.event_count) AS error_events,
  ROUND(
    1.0 * SUM(e.event_count) / NULLIF(MAX(r.sessions), 0), 4
  ) AS errors_per_session
FROM error_rollups e
JOIN releases r ON r.id = e.release_id
WHERE e.environment = 'production'
GROUP BY r.version
ORDER BY MIN(r.released_at) DESC
LIMIT 15;

Pitfalls

Where this metric applies

• Sentry + Metabase — error rollups with release and session context
• Raygun + Metabase — error groups and RUM sessions
• Prometheus + Metabase — request/error counters per service
• Cloudflare + Metabase — edge 4xx/5xx rates by zone

Related

Metrics

Dashboards

FAQ