Observability export

Planned Nothing on this page is built. Today the hub exports no metrics.

The direction, decided on 23 October 2026: the hub prepares metrics for your own observability tools, opt-in, with nothing sent anywhere unless an owner configures a destination. The release channel stays the only connection the hub makes by default.

What is planned

• The hub’s metrics, pushed over OTLP/HTTP every 60 seconds to an endpoint you configure (Grafana Cloud or any OTLP receiver), with Prometheus remote_write as the fallback. Push, not scrape: connections go outward, as everywhere in the hub.
  • Series such as each environment’s state (the Twin Row’s colours), twins, active agents, pending approvals and the oldest one’s age, offered releases, rollouts, the backup’s lag, and the hub’s memory against its limit.
  • Never exported: secrets, tokens, request bodies, IP addresses, personal names, approval contents. Environment and twin names are exported by default; a setting replaces them with stable hashes.
  • Destinations follow the webhook’s rules: HTTPS only, public addresses only, no redirects.
• A Grafana dashboard and example alert rules: an environment down for 5 minutes, an approval older than 10 minutes, backup lag over 5 minutes, hub memory over 80 % of its limit.
• A second destination for your servers’ metrics and logs, set per environment by the hub. Twins never forward by default.
• A /metrics endpoint to scrape, once the hub’s admin plane speaks HTTPS.
• The audit as OTLP logs with long retention, as part of the planned business plan.