I sort of wound up with a formidable little fleet of sites.. tronswan, chomptron, swantron, mt, wrenchtron, jswan.dev. The DevX dawg in me naturally wanted to know if they were up.

So I wrote the dead simplest thing that could tell me. uptime-monitor is 234 lines in one file.. a GitHub Actions cron that pings every site every 5 minutes and writes uptime to a public gist. No VM, no terraform, $0. It only tips over when GitHub tips over. I’d build it the same way again.

But it can’t tell me whether a deploy worked. Every deploy hands me a green checkmark, and the checkmark just means the pipeline exited 0. A HEAD / coming back 200 can be a CDN handing you a cached page while the app behind it is face-down. A pinger falls for that too.

OTel, Backwards

Usually you instrument with OpenTelemetry so you can stare at dashboards after something breaks. I flipped it around. Right after a deploy, fire some synthetic traffic at the live URL, tag every request with a trace id, and check that the telemetry for that exact run shows up. If it does, the request made it all the way to the real app.. not a cache, not a stale instance. If it doesn’t, or it’s slow, the deploy fails.

watchtron is a prober, a tiny control plane on a free-tier GCE e2-micro, and a reusable workflow any repo can call after its deploy job:

verify:
  needs: deploy
  uses: swantron/watchtron/.github/workflows/verify.yml@main
  with:
    service: tronswan
    version: ${{ github.sha }}
  secrets:
    otlp_endpoint: ${{ secrets.WATCHTRON_OTLP_ENDPOINT }}
    token: ${{ secrets.WATCHTRON_TOKEN }}

The fleet is all over the place.. DigitalOcean, Cloud Run, GitHub Pages, Firebase, a self-hosted PDS.. and one workflow covers all of it. The two Express apps I actually own also report back their own span and git SHA, so watchtron can tie the response to the probe and confirm the build that answered is the one that just shipped. Green means this build is serving, not just something.

Two Tools, One Strip

I started bolting history and dashboards onto watchtron and caught myself rebuilding the uptime monitor, worse. So I quit. The pinger is the heartbeat, watchtron is the deploy gate. Instead of competing, they share a chart: one strip per site, green up, red down, little triangles where deploys landed.

watchtron dashboard: per-service uptime strips with verified-deploy markers

It’s up at watch.swantron.com, with a cleaner cut on tronswan.com/status.

The dashboard paid for itself right away: its staleness flag noticed watchtron’s own cron had gone quiet. GitHub disables scheduled workflows after ~60 days without a commit. The uptime monitor has a dirty keepalive for exactly this. watchtron’s cron.. did not.

Source: github.com/swantron/watchtron

Now the green checkmark actually means something.