Exporters

MoniGo publishes the same runtime metrics through two paths. Prometheus scrapes MoniGo; MoniGo pushes to an OpenTelemetry collector. Both report the same figures under the same names, so a scrape and a push cannot disagree.

Prometheus

The /metrics endpoint is registered automatically — there is nothing to configure. Point a scrape at it:

scrape_configs:
  - job_name: my-service
    static_configs:
      - targets: ['localhost:8080']

Exposed metrics:

Metric	Type
monigo_cpu_usage_percent	gauge
monigo_memory_usage_bytes	gauge
monigo_goroutines_count	gauge
monigo_disk_read_bytes_total	counter
monigo_disk_write_bytes_total	counter

Mounting the dashboard

/metrics sits at the root, not under the API base path. If you register MoniGo through RegisterDashboardHandlers or the Fiber handler, make sure you are on v1.7.0 or later — earlier releases returned HTTP 500 on that path under those handlers, which Prometheus reads as a target being down.

OpenTelemetry

monigoInstance := monigo.NewBuilder().
    WithServiceName("my-service").
    WithOTelEndpoint("otel-collector:4317").
    Build()

Metrics are pushed over OTLP/gRPC every 30 seconds, with the first export running at startup rather than waiting for the first tick.

Upgrade if you are below v1.7.0

Before v1.7.0, WithOTelEndpoint connected to the collector and then sent nothing at all — no error, no warning, and a log line reading OTel exporter initialized. If you configured OTel export on an earlier release and never saw data arrive, this is why.

Reading exporter status

The dashboard’s Exporters page reports both paths, each in its own terms.

They are deliberately not flattened into one pass/fail, because they do not mean the same thing:

	Prometheus	OpenTelemetry
Direction	pull — it scrapes MoniGo	push — MoniGo sends
Reports	last scrape, scrape count	last attempt, last success, consecutive failures, the transport error
Failure count	none	yes

Prometheus carries no failure count because a failed scrape fails at the collector — the server never learns of it. Showing 0 failures there would assert a health signal that does not exist.

A configured exporter that has not run yet is listed as idle rather than omitted: absent from the page is indistinguishable from not configured, which is the one distinction worth having when you are checking whether export works.

States

State	Meaning
ok	last attempt succeeded
retrying	failing now, has succeeded before
failing	failing and has never succeeded — usually a wrong endpoint or no route
idle / not scraped	configured, nothing has happened yet

API

GET /monigo/api/v1/exporters

{
  "exporters": [
    { "name": "prometheus", "kind": "pull", "state": "ok", "total": 53,
      "last_success": "2026-08-30T14:16:31Z" },
    { "name": "otel-otlp", "kind": "push", "state": "ok", "total": 27,
      "failures": 0, "consecutive_failures": 0,
      "last_success": "2026-08-30T14:16:14Z",
      "last_attempt": "2026-08-30T14:16:14Z" }
  ]
}

Push-only fields are omitted for pull exporters rather than sent as zeros, and unset timestamps are omitted rather than serialised as a zero time.