Metrics

Introduction

Each pool's Metrics tab shows admit, queue, and reject proof for that pool. Use it to answer whether the concurrency limit is doing work, which keys burned slots, and whether rejects were auth, overload, or another outcome.

Metrics is not billing. Request allowance, overage, and spend caps live on Billing.

What you see

Outcomes: ok, overloaded, queue timeout, auth/IP rejects, upstream errors, and related statuses from the access log
Admit vs queue: p50 / p99 queue wait, plus live queue depth from the pool
Top keys: by admitted traffic and by overload rejects (key labels only)
Lock vs proxy: split between lock and proxy traffic
Response cache: hit / miss / bypass when the pool cache feature tagged the request
Policy hits: requests tagged with a traffic policy name

History length matches your plan's log retention: Starter 1 day, Team 2 days, Business 7 days.

Team and budget rollups

The pools index and Capacity budget pages show last-24h admit/reject rates per pool, with links into that pool's Metrics tab.

Alerts and export

Configure email threshold alerts on the Metrics tab. You get an email when reject rate, queue p99, or overload count crosses your threshold. CSV export is available for the current range.

Metrics vs Billing