Uptime Stats

Per-worker request counters since worker boot. Public, no auth, no credits.

Returns:

• total_requests: every request the worker has served
• by_status_class: {2xx, 3xx, 4xx, 5xx}
• error_rate_pct: 5xx / total * 100 (server-error rate)
• success_rate_pct: 2xx / total * 100
• server_error_first_at / server_error_last_at: timestamps of first/last 5xx (null if none)
• worker_started_at + worker_uptime_seconds (matches api-info)

Reset on every rolling reload (per-worker, in-process). The 5- worker pool means a customer polling sees one worker's stats per poll. Over time the round-robin load balancer surfaces a representative sample.

For a cross-worker aggregate uptime % a future iter could push these counters to Redis on a periodic flush. For now the per- worker view is the meaningful signal: "is this worker healthy?"

Pairs with /v1/meta/sla (targets) + /v1/meta/incidents (record of misses). All three discoverable through /v1/meta/api-info.

Stability: this is /v1/meta/uptime version 0.1. Keys may evolve.