Traffic

Traefik exposes no per-request API and no traffic stream. The only near-real-time signal it publishes is its Prometheus /metrics endpoint — so that is what the Traffic section uses.

How the numbers are made

Kubo scrapes the text exposition format (with the same parser the kube-state-metrics module uses), aggregates the relevant cumulative counters per object —

• traefik_{entrypoint,router,service}_requests_total, plus the code label for errors,
• _request_duration_seconds_sum and _count,
• _open_connections,

— and then differences two consecutive scrapes. That difference is what produces:

• per-second request rate;
• error percentage;
• average latency (duration sum over count, across the window);
• live open connections (a gauge, so it is read directly).

A segmented toggle groups all of it by entry point, router or service — the same number seen from three angles: which port is busy, which route is busy, which backend is busy.

Because rates come from differencing, the first scrape after opening the screen shows counters only; the numbers appear once the second one lands.

Where the metrics live

The metrics endpoint defaults to <serverUrl>/metrics, but in most real deployments metrics are exposed on a separate entry point or port — so a connection can set a Metrics URL override.

If the endpoint 404s, the screen says what to do — enable metrics.prometheus, or set a Metrics URL — rather than showing a generic error. A missing optional feature should read like a missing optional feature.