Saltar al contenido
GuíasAutomatizaciónIntermedio

Home Server Monitoring: A Complete Guide

Monitor your homelab with Uptime Kuma, Prometheus and Grafana: what to track, sensible alert thresholds, SMART disk health, backup checks and how to avoid alert fatigue.

por HomeServersGuide TeamActualizado 27 de octubre de 20264 min de lectura

Why monitor your home server?

A home server quietly does its job until, one day, it doesn't. A disk fills up, a drive starts throwing errors, a container crashes at 3 a.m., a backup silently stops, or the whole box overheats in a summer heatwave. Monitoring turns those silent failures into early warnings so you can fix problems before they cause data loss or downtime. It also tells you when something is wrong but not yet broken, which is the cheapest time to fix it.

There are three complementary layers of monitoring:

  • Availability monitoring — is the service up and responding?
  • Metrics monitoring — how is it performing over time (CPU, RAM, disk, temperature, network)?
  • Job monitoring — did the nightly backup, update or scrub actually run?

Start simple: Uptime Kuma

If you only set up one tool, make it Uptime Kuma. It runs in a single container, has a friendly UI, and checks whether your services respond over HTTP, TCP, DNS, ping and more.

  • Add a monitor for each service (your NAS UI, reverse proxy, media server, DNS server).
  • Configure a notification channel — email, Telegram, Discord, ntfy and dozens more.
  • Use keyword checks so a page that loads but shows an error still alerts.
  • Publish a status page so the whole household can see what's up.
  • Use push monitors (heartbeats) for jobs: your backup script calls a URL when it finishes, and Kuma alerts you if it ever doesn't.

Within ten minutes you'll get a message the moment anything goes offline.

Go deeper: Prometheus + Grafana

For trends and hardware health, pair Prometheus (collects metrics) with Grafana (draws dashboards). The full walkthrough is in our Grafana and Prometheus tutorial. In short:

  1. Run node_exporter on each host to expose CPU, RAM, disk and temperature metrics.
  2. Run cAdvisor to expose per-container metrics.
  3. Point Prometheus at those exporters to scrape them every 15–30 seconds.
  4. Add Prometheus as a data source in Grafana and import a community dashboard (there are excellent ready-made ones for node_exporter and Docker).

What to actually watch

SignalWhy it mattersSensible alert
Disk spaceA full disk breaks databases and backupsBelow 15–20% free
Drive health (SMART)Reallocated or pending sectors mean a dying driveAny non-zero reallocated/pending count
TemperaturesHeat shortens hardware lifeCPU above ~85 °C, drives above ~50 °C sustained
Memory pressureConstant swapping slows everythingSwap in heavy use for 10+ minutes
Container restartsA restart loop signals a real bugMore than 3 restarts in an hour
BackupsA backup that silently fails protects nothingNo success in 24–48 hours
CertificatesExpired HTTPS breaks accessFewer than 14 days left
ZFS/RAID stateA degraded array is one failure from disasterAny state other than healthy

For a NAS, enable your software's own SMART tests and ZFS scrubs and make sure alerts reach you. See RAID explained and build your first NAS.

Set up alerts you'll actually see

Monitoring without alerting is just pretty graphs. Route alerts to a channel you check — a phone push via Telegram or ntfy works well. Some principles that prevent alert fatigue:

  • Alert on symptoms you can act on. "Disk 90% full" is useful; "CPU spiked for a second" is noise.
  • Require a duration, such as "for 5 minutes", before firing.
  • Keep alerts few. If everything alerts, you'll ignore all of it.
  • Escalate by severity. A failed backup deserves a loud push; a slow container deserves a daily summary.
  • Don't monitor from the machine being monitored. If the server goes down, so does its alerting. Run a second Uptime Kuma on a different device or use an external heartbeat service for critical checks.

Monitor power and the UPS

If you have a UPS, monitor its battery state with Network UPS Tools (NUT) so you know when it is on battery or needs replacing. A power cut should send you an alert as well as a clean shutdown.

Review regularly

Once a month, glance at the dashboards. Slow trends — a pool filling by 1% a week, a drive getting hotter — are exactly what monitoring is good at revealing. Pair this review with your update and backup routine from the beginner's guide.

Next steps

Once the basics are running, add log aggregation (Loki) and a status dashboard on a spare screen if you enjoy it. But don't over-engineer: for most homelabs, Uptime Kuma plus one Grafana dashboard covers 95% of what you need. Make sure your monitoring is also protected — keep dashboards on your internal network or behind a VPN, as described in the security checklist.

Artículos relacionados

TutorialesAutomatizaciónIntermedio

How to Set Up Grafana and Prometheus

Stand up a metrics stack with Prometheus, node_exporter, cAdvisor and Grafana to visualize your home server's health, with retention, security and alerting tips.

4 min de lectura
TutorialesAutomatizaciónPrincipiante

How to Install Uptime Kuma

Install Uptime Kuma with Docker to monitor your homelab services, get alerts on your phone, track backup jobs with push monitors and publish a status page.

5 min de lectura
TutorialesAutomatizaciónIntermedio

How to Install Home Assistant

Install Home Assistant for private, local-first smart home automation: HAOS vs Container, Docker Compose, integrations, Zigbee and MQTT, automations, backups and remote access.

5 min de lectura