Plex Server Disk Space Monitoring with Prometheus and Grafana
Updated November 19, 2026 · PlexBeam
Why Monitor Disk Space?
A Plex server that runs out of disk space fails in unpredictable ways. The Plex database (SQLite) corrupts when it cannot write, transcoding jobs fail silently, library scans hang indefinitely, and new recordings from the DVR feature simply vanish. By the time you notice, you have already lost data.
Automated monitoring catches the problem before it becomes an emergency. A proper setup tracks disk usage trends over time, predicts when you will run out of space, and sends you a push notification when usage crosses a threshold — all without you having to log in and check manually.
Architecture Overview
The monitoring stack has three layers:
- Collection — Prometheus
node_exporterruns on the Plex server and exposes filesystem metrics at:9100/metrics - Storage and querying — Prometheus scrapes node_exporter every 15 seconds and stores the time-series data
- Visualization and alerting — Grafana reads from Prometheus to render dashboards; Alertmanager (or Grafana's built-in alerts) triggers notifications when thresholds are crossed
All three components run as lightweight services. On a typical Plex server, the entire monitoring stack uses under 200 MB of RAM and negligible CPU.
Installing node_exporter
Node_exporter is a single binary with no dependencies. Download the latest release from GitHub, extract it, and run it. On Linux:
wget https://github.com/prometheus/node_exporter/releases/download/v1.8.2/node_exporter-1.8.2.linux-amd64.tar.gz tar xzf node_exporter-*.tar.gz sudo cp node_exporter-*/node_exporter /usr/local/bin/ sudo useradd -rs /bin/false node_exporter
Create a systemd service so it starts on boot and restarts on failure. The key flags are --collector.filesystem (enabled by default) and --collector.diskstats for I/O metrics. Bind to localhost:9100 if Prometheus runs on the same machine, or to 0.0.0.0:9100 if it is on a separate host (firewall port 9100 to only allow the Prometheus server).
Configuring Prometheus
Add a scrape target in prometheus.yml pointing to node_exporter. The scrape interval of 15 seconds is the default and works well for disk metrics (disk space changes slowly). Set a retention period of 15-30 days — longer retention is unnecessary for disk monitoring and wastes storage.
The key metrics for disk space monitoring are:
node_filesystem_avail_bytes— available space in bytes (what a non-root user can actually use)node_filesystem_size_bytes— total filesystem sizenode_filesystem_free_bytes— free space including reserved blocks (slightly more than avail_bytes)node_disk_io_time_seconds_total— cumulative I/O time per disk (high values indicate a disk bottleneck)
Grafana Dashboard Setup
Grafana connects to Prometheus as a data source. Create a dashboard with these panels:
Panel 1: Disk Usage Gauge
A gauge showing current usage percentage per mount point. Query: 100 - (node_filesystem_avail_bytes{mountpoint="/"} / node_filesystem_size_bytes{mountpoint="/"} * 100). Set thresholds at 70% (yellow), 85% (orange), and 95% (red). Repeat the panel for each mount point (media drives, config drive, transcode drive).
Panel 2: Usage Over Time
A time-series graph showing node_filesystem_avail_bytes over the past 30 days. This reveals the consumption trend: a steadily declining line means you are adding media faster than you are cleaning up. The slope tells you exactly when you will hit zero.
Panel 3: Predicted Full Date
Use Prometheus's predict_linear() function to forecast when a filesystem will be full. Query: predict_linear(node_filesystem_avail_bytes{mountpoint="/media"}[7d], 86400 * 30) predicts available space 30 days from now. Display this as a stat panel showing the estimated days until full.
Panel 4: Disk I/O
A graph of rate(node_disk_io_time_seconds_total[5m]) per disk. Values above 0.9 (90% busy) indicate the disk is saturated, which causes Plex to buffer during playback and slows library scans. This is especially useful for detecting a failing drive that is spending excessive time on read retries.
Setting Up Alerts
Alerts trigger actions when metrics cross thresholds. You can configure alerts in Grafana directly (simpler) or use Prometheus Alertmanager (more powerful routing and deduplication).
Grafana Alert Rules
In each dashboard panel, click the "Alert" tab and define conditions. For disk space, a reasonable rule is: fire when available space drops below 10% for more than 5 minutes. The 5-minute window prevents false alarms from temporary spikes (large transcode jobs can briefly consume space and release it).
Notification Channels
Grafana supports many notification targets out of the box:
- ntfy — self-hosted push notifications to your phone. Set up a webhook contact point in Grafana pointing to
http://localhost:2586/plex-alerts(or your ntfy server URL). Subscribe to theplex-alertstopic in the ntfy app on your phone. - Gotify — similar to ntfy, another self-hosted push notification server. Use the Gotify webhook URL with your application token.
- Email — configure Grafana's SMTP settings in
grafana.iniunder[smtp]. Works with any mail server including Gmail's SMTP relay. - Discord/Slack — use the built-in Discord or Slack integration with a webhook URL from your server.
Alert Escalation Strategy
A single alert is not enough. Set up a tiered system:
- Warning at 80% usage — push notification. You have time to clean up or add storage.
- Critical at 90% usage — push notification + email. Take action within 24-48 hours.
- Emergency at 95% usage — push notification + email + Plex automatically pauses DVR recording and disables the media optimizer to prevent further writes.
For the emergency tier, you can trigger automated remediation with a webhook that calls a script. The script might delete old transcode cache files (find /transcode -mtime +1 -delete), clear the Plex photo cache, or empty the trash in all libraries.
Docker Compose for the Full Stack
If your Plex server runs in Docker, deploy the entire monitoring stack alongside it in the same Compose file. Node_exporter, Prometheus, and Grafana each have official Docker images. Mount the host's filesystem into node_exporter's container (--path.rootfs=/host) so it reports the host's disk metrics rather than the container's overlay filesystem.
Total resource overhead for the monitoring stack: ~150 MB RAM, ~1 GB disk for 30 days of metric retention, and effectively zero CPU on a modern processor.
Never Run Out of Space Again
PlexBeam covers monitoring, automation, and infrastructure for serious Plex deployments.
Browse All Guides →