Prometheus metrics: /metrics on the hub (and runners) issue #212 closed: completed andymccall opened this · 1 comment
referenced by 6 changes on 18 Oct 2026
8a48dd830178 b9b74d997cec a6d824c5ebf5 2c47eecfeab6 85ec74fd82d8 0843f23ec51c
andymccall
Board tidy (2026-10-19): the hub's /metrics is live (first slice 2026-10-18: build, uptime, memory, requests by route, fetch/push, Flow queue and job times; token-protected), scraped on GCP into Managed Prometheus with the Grafana dashboard (oxygenhub-gcp issue #12). Left: the runners' own /metrics (claims, job durations by tier, cache hits, image fetches), which overlaps oxygenflow#35. Moving to Backlog until then.
closed by change 9e8cbfc3ad00 ·
andymccall referenced by change ab2d8b40f9f4 on main ·
For monitoring oxygenhub.io (andymccall/oxygenhub-gcp: Prometheus and Grafana). The hub serves Prometheus metrics at /metrics, private (a bearer token or the internal network only, never through Cloudflare): HTTP requests by route and status with latency histograms, slow requests, pushes and fetches (count, bytes, duration), Flow (queued, running and waiting jobs, run results, queue wait, runner heartbeats), backups (last run, result, age), Diffusion (bumps, held, pin index repairs from issue #210), mail sent and failed, notifications delivered, database pool and query latency, storage (repository bytes, GCS store latency), process (memory, open files, uptime, version). The flow runner exposes its own: jobs, VM boots, image fetches.