Skip to main content
The Prometheus metrics endpoint now exposes a single metric that breaks queue time down by state. mq_pr_state_duration_1h_seconds reports the time PRs spent in each individual queue state on a state label, covering not_ready, pending, testing, tests_passed, pending_failure, and the terminal states. One metric replaces the need to add a new series each time you want to chart another phase of the queue.
  • Decompose Time in queue by state. Select state=~"pending|testing|tests_passed|pending_failure" to see exactly which phase is driving your queue time.
  • Percentile gauges are now documented. Every duration histogram on the endpoint also emits _p50, _p75, _p90, _p95, and _p99 gauges you can read directly, without a histogram_quantile() over _bucket series.
  • Existing metrics are unchanged. mq_pr_wait_duration_1h_seconds and mq_pr_test_duration_1h_seconds keep emitting the same values, so current dashboards and alerts continue to work.
See Per-state durations for the full label reference, PromQL examples, and guidance on when to prefer the new metric over the existing ones.