Overview

Scraparr is a Prometheus exporter for the *arr suite. It talks to the API of each app you point it at — Sonarr, Radarr, Prowlarr, Bazarr, Readarr, Whisparr, Jellyseerr, Overseerr and Jellyfin — collects what that app knows about itself, and republishes it as Prometheus metrics on a single endpoint at :7100/metrics. Prometheus scrapes that endpoint like any other exporter, and from there the usual stack follows: graphs, alerts and dashboards that sit next to the rest of your homelab metrics.

What makes it useful is the translation step. The *arr apps expose rich APIs but nothing Prometheus can scrape, so a library’s queue, its missing episodes, its disk usage and its quality mix stay locked inside nine separate web interfaces. Scraparr turns each of those into a gauge — sonarr_queue_count, sonarr_missing_episodes, sonarr_available_disk_size, and the equivalent per service — with an alias label so several instances of the same app can be told apart.

Key Features

• Nine connectors: Sonarr, Radarr, Prowlarr, Bazarr, Readarr, Whisparr, Jellyseerr, Overseerr and Jellyfin, each with its own metric set.
• Queue health as metrics: queue size plus separate error and warning counters, so a stuck download is an alert rather than something you notice a week later.
• Library and disk numbers: item counts, monitored versus unmonitored, total, free and available disk size per configured path.
• Breakdowns: items per quality profile and per genre, so you can see what your library is actually made of.
• Missing-item gauges: missing episodes and equivalent per-service counters — the raw material for a “still missing” alert.
• Scrape introspection: last-scrape timestamp and scrape duration per alias, so you can alert on the exporter itself.
• Multiple instances: each service can be a list of instances with an alias; the label keeps their metrics from overwriting each other.
• Per-service tuning: optional interval (default 30 seconds), api_version, and a detailed mode for per-item data.
• Config file or environment variables: a mounted YAML config, or env vars for the simple single-instance case.
• Exporter-side auth: optional HTTP basic auth or bearer token on the metrics endpoint.
• Grafana dashboards included: the main dashboard is published as Grafana.com dashboard 22934, with more in the repository’s dashboards folder.

Why Use It

Monitoring the *arr stack usually means monitoring everything around it: disks, containers, the download client — and then opening six web UIs to answer “is anything stuck?”. Scraparr closes that gap with one small process and one scrape job. Once the numbers are in Prometheus, the questions become queries: which instance has queue errors right now, which series are still missing episodes, which path is running out of space, how the quality mix shifted after a profile change.

It is deliberately thin. There is no database, no UI to learn and no agent on each app — just a Python service that polls APIs you already expose, so adding it costs one container and one stanza of scrape configuration.

Use Cases

• One dashboard for the whole stack: queue, library and disk panels per app on a single Grafana board instead of nine browser tabs.
• Alerting that matters: page on queue errors, a scrape that stopped running, or a mount below a free-space threshold.
• Multi-instance setups: separate 4K and 1080p Sonarrs, or several Radarrs, kept apart with aliases in one config.
• Capacity planning: track disk growth per path and per quality tier over months rather than guessing.
• Library hygiene: watch missing and unmonitored counts drift, and spot a season that never completed.

Besides the Docker image, a community-maintained Helm chart for Kubernetes and an Unraid template exist, and it can run from source with Python 3 by installing requirements.txt and running python -m scraparr.scraparr. Configuration is one YAML file mounted at /scraparr/config/config.yaml; inside Docker, reach host-side services by IP or via the host.docker.internal:host-gateway extra host.

Licence

Free and open source — GPL-3.0. Upstream development lives on Codeberg; the GitHub repository is a mirror.