← Beacon for Grafana

What Beacon actually watches

The real homelab behind the polite little menu bar icon.

Beacon is only useful if Grafana has something worth saying, so here is the actual setup behind mine. It covers the network, storage, containers, a couple of social servers, a 24/7 radio stream, some media gear, and the little AI corner of the house where GPUs turn electricity into invoices.

Grafana watches the lot. Alerts only fire when something genuinely needs me, because life is short and I refuse to be paged by my own hobby unless the hobby has earned it.

~13
dashboards
36
alert rules
8
domains

Network & internet

The boring foundation, which naturally becomes fascinating the second it stops working. If the line drops or WiFi starts sulking, I want Beacon to say so before someone asks why Netflix has become a slideshow.

  • WAN drops, gateway latency, jitter, and packet loss
  • Broadband speed sagging below the contractually optimistic number
  • Per-device WiFi signal, and the odd top-talker hogging the link

Certificates & reachability

The classic self-hosting humiliation: a certificate expires, every browser puts on a hazmat suit, and suddenly your clever setup looks like malware.

  • Wildcard TLS cert expiry, with a 21-day warning while there is still dignity left
  • ACME renewal loops that are failing quietly instead of renewing loudly

Storage (TrueNAS / ZFS)

This is where the stuff I cannot casually re-download lives, so it gets the closest watch, including a watch on the watcher. Paranoid? Fine. Correct? Also fine.

  • A ZFS pool dropping out of ONLINE, immediately
  • A runaway process chewing reads/writes for half an hour straight
  • The storage exporter going down, because alerts need adult supervision too

Containers (Docker)

Every service gets a chance to report that it is feeling unwell before it collapses onto the floor. Basic manners.

  • A container unhealthy for more than five minutes
  • Memory creeping toward its limit, before the OOM killer steps in
  • Crash loops, also known as optimism with a restart policy

Logs (across everything)

The logs read themselves until a pattern changes enough to deserve a human. A beautiful arrangement, mainly because the human is me.

  • Sustained error floods instead of one dramatic stack trace having a moment
  • Error-rate spikes at 5x each service's rolling baseline, so noisy services do not win by being noisy

Databases (PostgreSQL)

Connections, cache, replication: all wonderfully boring until they become a documentary about consequences.

  • Replication falling behind
  • Deadlocks and lock pile-ups, the database version of everyone standing in a doorway
  • Unhealthy growth in size, temp files or WAL

Social & fediverse

I host my own Mastodon and a Bluesky PDS, which means I am also the ops team, support desk, and unlucky intern. Beacon helps me notice trouble before my followers do.

  • Instance, web or streaming going down
  • Background jobs dying, failing or stuck retrying
  • Federation peers disappearing, or the database putting on more than 100 MB a day

Live radio & media

A 24/7 stream, Plex, a GPU-backed AI gateway, and the rest of the supposedly fun stuff that becomes less fun when it quietly cooks itself.

  • The Icecast source disconnecting
  • Pipeline latency climbing, or a “clock spiral” runaway drift
  • GPU VRAM and power draw getting spicy before the AI gateway impersonates a space heater

All of that collapses into one quiet menu bar icon, right up until one of those checks decides it would like to ruin my afternoon.

← Back to Beacon