Decision
JMX + RCON together
over a single exporter
JMX gives deep JVM metrics; RCON gives live control. Together they cover "what’s happening" and "do something about it".
IoT · Game server
Minecraft Monitoring
A game server treated like production — 95.2% uptime, TPS-drop alerts, and a 10-command Discord bot. The live stats on this site come from here.
95.2%uptime
10Discord commands
real-timeTPS · heap · players
- PaperMC
- JMX
- RCON
- Discord.py
- Prometheus
- Grafana
Friends kept saying "it’s lagging" with no data
A Minecraft server is a real-time system with real users, and "is it lagging?" deserves a real answer. I wanted tick-rate, heap, and player metrics scraped continuously, alerts when ticks drop, and a way to check or fix things without SSH.
It also became the honest heartbeat for this whole site: a cron publishes a snapshot that the home page and dashboard read, so the "live" claim is backed by the same feed I use to run the server.
Export, scrape, dashboard, control, publish
JMX and RCON expose the server’s internals; Prometheus scrapes them; Grafana visualises; a Discord bot alerts and controls; and a cron publishes a JSON snapshot for this site to read.
Game
PaperMC
Exporters
JMX metricsRCON control
Signals
Prometheus
Dashboards
Grafana (20 panels)
Ops
Discord bot (10 cmds)TPS-drop alerts
Publish
cron → JSONthis site’s live strip
What I chose, and why
Decision
A published snapshot
over exposing Prometheus publicly
A cron writes a small JSON the site reads — no public scrape endpoint, and an honest staleness timestamp comes for free.
Decision
A Discord bot
over a web control panel
Ten slash commands (status, players, tps…) put ops where the players already are, with no extra service to host.
The results
95.2%Uptimelive, 7-day average
10Bot commands
15sScrape interval
20Grafana panels
The same feed powers the live strip on the home page and the dashboard.