Skip to content
New here? Rent a GPU by the second
Kracht
Sign inSign up

Status

Monitoring is not connected yet

This page says what is observed today, what is not, and how to reach us when something is wrong.

There is no live status to report

Nothing on this page is a measurement, because nothing collects one yet. It used to show uptime percentages and regional health as though they had been observed. They were written by hand, and a page whose whole job is to be live cannot make that honest by labelling it.

unknown: no trustworthy information exists, source no collector is running

What is observed today

These run continuously and act on what they see. None of them is published, so none of them can be charted here yet.

  • Host heartbeats

    Every provider machine reports in on a schedule. When one goes silent the worker marks it offline and rescues the rentals stranded on it, so a dead box cannot keep billing.

    services/worker · detect_dead_hosts

  • Metered settlement

    Usage is metered per second and settled to the microeuro, with the split applied at exact integers. Every posting is idempotent, so a retried heartbeat cannot bill twice.

    services/worker · settle_finished_instances

  • Spend limits

    Account spend caps are enforced by the worker rather than by the console, so a runaway job stops even when nobody is watching a screen.

    services/worker · enforce_spend_limits

  • A health endpoint and raw metrics

    The control plane answers /healthz and renders Prometheus metrics derived from database state. Nothing collects them yet, which is exactly why there is no history on this page.

    services/control-plane · api/health, services/metrics

What is not observed

In the same type size as the section above, because a gap stated quietly is a gap hidden.

  • Uptime, for any component. There is no probe and no history, so any percentage here would have been typed by a person.
  • Latency, per region or otherwise. Nothing measures round-trip time from outside our own network.
  • Regional health. Two of the four regions this page used to show as monitored have never had a host.
  • Incident history. Incidents have not been recorded in a form anyone could publish.

What this page will show when collection lands

A preview of the layout, drawn with invented days so the shape is reviewable before there is anything real to put in it. None of these bars is an observation, no component below has been measured, and the place a percentage will go says so instead of showing one.

illustrative: illustrative only, not observed

Control plane

not measured

The API that places work and answers /healthz.

Layout preview · no incidents recorded · 90 days

Control plane: a layout preview over 90 days. Nothing here has been measured.

90 days agotoday

Settlement worker

not measured

Metering, the split, and the dead-host rescue.

Layout preview · no incidents recorded · 90 days

Settlement worker: a layout preview over 90 days. Nothing here has been measured.

90 days agotoday

Host fleet

not measured

Provider machines reporting in on schedule.

Layout preview · no incidents recorded · 90 days

Host fleet: a layout preview over 90 days. Nothing here has been measured.

90 days agotoday

No percentage and no incident count is printed, because a figure computed from invented bars is still an invented figure. When collection lands, the same rows print the real ones.

If something is broken right now

Until there is a page that can tell you, a person can. Include the instance id and roughly when it started, and you will get a reply from someone who can read the ledger.

Contact supportWhat has shipped recently

When collection is wired up, this page will show real component states, real regional state, incidents with timestamps, and the method behind each figure. It will not show any of those things before then.