NAS monitoring: watching disks, services and logs at a glance

Why monitor your NAS

On a machine running 24/7, disks age, services crash silently, and Docker logs pile up unread — until a disk starts showing wear or a container restarts in a loop with nobody noticing.

Four tools cover the need, each with a precise role and no overlap, all in Docker on an isolated monitoring_net network, with configs in /mnt/data/apps/monitoring/<service>/:

• Scrutiny (port 9998): SMART disk health
• Uptime Kuma (port 3001): service availability
• Dozzle (port 9999): live Docker logs
• Homepage (port 3010): unified dashboard

Scrutiny: disk health

The critical tool. Scrutiny collects SMART data from every disk and displays it with historical trends. The metrics that matter:

• Temperature: an overheating disk wears out faster
• Reallocated Sectors: reallocated sectors signal physical degradation
• Current Pending Sectors: sectors awaiting reallocation, an imminent problem
• Read/Write errors: rising error counts announce a failure

Scrutiny has two parts: a collector that polls the disks periodically and a web UI for results. The Docker config is unusual, since it needs direct disk access:

# docker-compose.yml (monitoring excerpt)
services:
  scrutiny:
    image: ghcr.io/analogj/scrutiny:master-omnibus
    container_name: scrutiny
    restart: unless-stopped
    ports:
      - '9998:8080'
    volumes:
      - /mnt/data/apps/monitoring/scrutiny/config:/opt/scrutiny/config
      - /mnt/data/apps/monitoring/scrutiny/influxdb:/opt/scrutiny/influxdb
      - /run/udev:/run/udev:ro
    devices:
      - /dev/sda
      - /dev/sdb
      - /dev/sdc
      - /dev/sdd
    cap_add:
      - SYS_RAWIO
    networks:
      - monitoring_net

cap_add: SYS_RAWIO is mandatory: without it, the collector can't query disks over SMART. Mounting /run/udev read-only allows disk identification.

Uptime Kuma: service availability

Uptime Kuma covers HTTP, HTTPS, TCP, Ping and DNS. Every NAS service gets its own monitor.

uptime-kuma:
  image: louislam/uptime-kuma:1
  container_name: uptime-kuma
  restart: unless-stopped
  ports:
    - '3001:3001'
  volumes:
    - /mnt/data/apps/monitoring/uptime-kuma/data:/app/data
  networks:
    - monitoring_net

What's watched: HTTP checks on Jellyfin, Immich, Home Assistant, Sonarr, Radarr; TCP checks on Samba (445), SSH (22), NFS (2049); and pings on the NAS itself, the router and the switches.

Beyond outright failures, the latency graphs are valuable: gradual degradation often reveals a disk, network or RAM problem before the service actually goes down.

Notifications go out via email, Telegram, Discord, Slack, Gotify or ntfy. You can also create public or internal status pages to share service state.

Dozzle: logs in real time

When a container misbehaves, the first thing to check is its logs. Dozzle streams them in a web UI, with no SSH and no docker logs.

dozzle:
  image: amir20/dozzle:latest
  container_name: dozzle
  restart: unless-stopped
  ports:
    - '9999:8080'
  volumes:
    - /var/run/docker.sock:/var/run/docker.sock:ro
  networks:
    - monitoring_net

Dozzle is deliberately minimal: no database, no persistence, no background collection. It reads the Docker socket and displays, which keeps it at a few MB of RAM. Day to day: search and filtering within a container's logs, a multi-container view to correlate events, auto-scroll with pause when you scroll back, and regex filtering.

Mounting the socket ro is enough: Dozzle reads, it doesn't control containers.

Homepage: the unified dashboard

With around twenty services, remembering ports gets tedious. Homepage provides a YAML-configurable dashboard.

homepage:
  image: ghcr.io/gethomepage/homepage:latest
  container_name: homepage
  restart: unless-stopped
  ports:
    - '3010:3000'
  volumes:
    - /mnt/data/apps/monitoring/homepage/config:/app/config
    - /var/run/docker.sock:/var/run/docker.sock:ro
  networks:
    - monitoring_net

An excerpt from services.yaml:

# /mnt/data/apps/monitoring/homepage/config/services.yaml
- Media:
    - Jellyfin:
        href: http://192.168.1.50:8096
        icon: jellyfin.svg
        description: Media server
        widget:
          type: jellyfin
          url: http://192.168.1.50:8096
          key: '{{HOMEPAGE_VAR_JELLYFIN_KEY}}'
    - Immich:
        href: http://192.168.1.50:2283
        icon: immich.svg
        description: Photos and videos

- Downloads:
    - qBittorrent:
        href: http://192.168.1.50:8080
        icon: qbittorrent.svg
        description: BitTorrent client
        widget:
          type: qbittorrent
          url: http://192.168.1.50:8080
          username: admin
          password: '{{HOMEPAGE_VAR_QBIT_PASSWORD}}'
    - Sonarr:
        href: http://192.168.1.50:8989
        icon: sonarr.svg
        description: TV show management

- Monitoring:
    - Scrutiny:
        href: http://192.168.1.50:9998
        icon: scrutiny.svg
        description: Disk health
    - Uptime Kuma:
        href: http://192.168.1.50:3001
        icon: uptime-kuma.svg
        description: Service availability

Widgets enrich the dashboard: Jellyfin shows active sessions, qBittorrent shows ongoing downloads.

The isolated Docker network

networks:
  monitoring_net:
    name: monitoring_net
    driver: bridge

Isolation buys two things: the monitoring containers talk to each other without going through published ports, and separation from the other stacks (media, downloads, home automation) stays clean.

Cockpit: the system complement

Alongside, Cockpit (port 9090) installed natively on Debian covers what the Docker tools can't see: CPU, RAM, swap, disk space, systemd services, web terminal.

sudo apt install cockpit

Cockpit for the host system, the four Docker tools for services and disks.

Verdict

Monitoring isn't a luxury on a NAS, it's insurance: Scrutiny flags an abnormal temperature before it turns critical, Uptime Kuma warns within seconds when a service goes down, Dozzle saves considerable debugging time, and Homepage acts as the single entry point. Four lightweight tools deployed in minutes through Docker Compose.