Skip to content

fix: configure histogram buckets for P99 latency dashboard#18

Merged
yurushao merged 1 commit intomainfrom
fix/histogram-buckets
Feb 21, 2026
Merged

fix: configure histogram buckets for P99 latency dashboard#18
yurushao merged 1 commit intomainfrom
fix/histogram-buckets

Conversation

@yurushao
Copy link
Copy Markdown
Member

Summary

  • The Prometheus exporter was defaulting to summary-style metrics (quantiles) instead of true histograms with _bucket labels
  • The Grafana P99 Latency panel uses histogram_quantile() which requires _bucket/le metrics, causing "No data"
  • Added explicit bucket boundaries (1ms to 10s) to PrometheusBuilder so histograms export correctly

Test plan

  • cargo build passes
  • All 38 tests pass
  • After deploy, verify rpc_request_duration_seconds_bucket appears in /metrics output
  • Confirm P99 Latency panel populates in Grafana within ~2 minutes of scraping

🤖 Generated with Claude Code

The Prometheus exporter defaults to emitting summaries (quantile-based)
instead of true histograms with _bucket metrics. The Grafana dashboard
uses histogram_quantile() which requires _bucket/le labels, resulting
in "No data" for the P99 Latency panel.

Add explicit bucket boundaries to PrometheusBuilder so histograms are
exported with _bucket, _sum, and _count suffixes.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@yurushao yurushao merged commit ede1b1c into main Feb 21, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant