Skip to content

prometheus

Configuration for Prometheus, a metrics collector and time series database. Each job of prometheus.jobs scrapes a static list of host:port targets over HTTP (/metrics by default) every scrape_interval, and the samples are kept for retention in a named volume. Prometheus also scrapes itself. The generated /home/u_prometheus/config/prometheus.yml is bind-mounted read-only into the container: edit it as u_prometheus and Prometheus re-reads it within 30 seconds, no restart needed.

Its purpose in this repository is to feed the Homepage dashboard: a prometheus entry in homepage.services shows the targets up and down, and an entry of homepage.custom_services with a prometheusmetric widget displays the result of any PromQL query, for any application exposing metrics (see there).

Note

Prometheus has no authentication and its web interface lets anyone query the collected data. Homepage reaches it on host.containers.internal:<http_port> without any Caddy site, so either do not publish it at all or list its site in caddy.private_sites, so that only the local network gets an answer, or in caddy.protected_sites, as in the example metaconfig.yaml, so that clients outside the LAN log in first.

prometheus.image

The image of Prometheus to deploy.

  • Type: String
  • Example: docker.io/prom/prometheus:v3

prometheus.http_port

Port to listen for HTTP requests (web interface and API).

  • Type: Integer
  • Example: 3013

prometheus.retention

How long samples are kept before being deleted from the data volume.

  • Type: String (Prometheus duration: y, w, d, h, m, s)
  • Example: 30d

prometheus.scrape_interval

Interval between two scrapes of a target, for the jobs that do not set their own.

  • Type: String (Prometheus duration)
  • Example: 15s

prometheus.jobs

Scrape jobs, one per monitored application. Each job has a name (the job label of its series) and a list of targets (host:port). An optional labels map is attached to every series of the job, and any other key is copied as-is into the generated scrape_config (scrape_interval, metrics_path, scheme, basic_auth, params..., see the Prometheus configuration reference).

  • Type: List of maps

  • Example:

    jobs:
  - name: vllm
    targets:
      - 192.168.0.27:8000
    scrape_interval: 5s