RackMon — deployment kit & SNMP temperature reference

Everything you need to run RackMon on your own hardware, plus a vendor-by-vendor reference of the SNMP OIDs that actually return rack temperature.

RackMon is a self-hosted monitor built around one job: is anything in my racks getting too hot, and can I prove it later? Live per-U thermal heatmaps, SNMP v2c/v3 polling, alerting that doesn't lie to you, availability/SLA reports and a public status page — in one container.

Live demo · read-only, real data, no signup — viewer@rackmon.io / viewer12345

_{Heads up on licensing: RackMon itself is not open source. This repository holds the deployment tooling, configuration and reference data, which are MIT-licensed and yours to fork. The application ships as a container image. The Community edition is free forever with unlimited devices and metrics; a paid Pro tier adds the client-facing features. Details in LICENSING.md — no surprises.}

---

Quickest start

docker run -d --name rackmon -p 8000:8000 \
  -v rackmon_data:/data \
  -e RACKMON_JWT_SECRET="$(openssl rand -hex 32)" \
  -e RACKMON_MASTER_KEY="$(openssl rand -hex 32)" \
  -e RACKMON_DATABASE_URL="sqlite:////data/rackmon.db" \
  rackmon/rackmon:latest

Open http://localhost:8000. That's it — SQLite, no external database, nothing to configure. Point it at real gear when you're ready, or leave the simulator on to look around first.

For anything beyond a single node, use the compose file below.

What's in here

Path	What it is
docker-compose.yml	Production single-node stack: RackMon + TimescaleDB + Caddy (TLS)
docker-compose.sqlite.yml	Minimal one-container setup, SQLite, no reverse proxy
.env.example	Every environment variable, documented, with safe defaults
snmp-profiles/	The OID reference — real temperature OIDs per vendor
alerts/	Example alerting rules with sane thresholds and hold-down windows
unraid/	Unraid Community Applications template
helm/	Helm chart for Kubernetes

The SNMP temperature reference

This is the part worth bookmarking even if you never run RackMon.

Nearly every guide about rack temperature says "poll your vendor's OID" and then doesn't tell you which one. snmp-profiles/ has them, per vendor, with the two things that actually trip people up documented alongside: scaling (a value of 275 meaning 27.5 °C) and units (a card quietly reporting °F).

Start with the vendor-neutral standard — a surprising amount of modern gear implements it:

# every sensor value the device exposes (ENTITY-SENSOR-MIB, RFC 3433)
snmpwalk -v2c -c public 10.0.0.10 1.3.6.1.2.1.99.1.1.1.4

# the human-readable names, so you know which index is the inlet
snmpwalk -v2c -c public 10.0.0.10 1.3.6.1.2.1.47.1.1.1.1.2

Covered so far: the RFC 3433 standard, Cisco, Juniper, Dell iDRAC, HPE iLO, APC/Schneider NMC.

Corrections and additions are very welcome — if you have gear that isn't listed, or an OID here is wrong on your firmware, please open an issue. Model- and firmware-specific quirks are exactly the knowledge that's hard to find and easy to share.

Measure the right thing

The most common mistake is monitoring the CPU temperature the server reports. That tells you how hot the chip is, not how hot the air reaching it is — and the second number is the one that warns you early.

• Inlet temperature — cold-aisle air at the front of the rack. This is the metric that matters; it's what vendors define their limits against, and it's what moves when cooling fails.
• Exhaust temperature — useful as a delta against inlet. A widening gap usually means dirty fans, clogged filters, or a box working harder than it should.
• Internal/CPU — fine as confirmation, but it lags. By the time it climbs, you're already late.

A sane starting point, measured at the rack inlet: normal up to 27 °C, warn 28–32 °C, critical above 32 °C (ASHRAE class A1–A2 recommends an 18–27 °C inlet range). Check your own vendors' limits and treat these as a start, not gospel.

Alerting that stays trustworthy

A thermal alert that cries wolf gets muted within two weeks, and then you have no monitoring at all — you have noise. Three rules do most of the work, and they're the defaults in alerts/examples.yaml:

• Hold-down windows — don't fire on the first poll over threshold. Heat is a slow physical quantity; a lone spike is almost always a bad read.
• Deduplication — when cooling fails, twenty devices cross the line at once. That's one incident, not twenty pages.
• Never fake a recovery — a missed poll is missing data, not a return to normal. Plenty of tools mark "recovered" when they've actually gone blind, which is precisely when you need them.

Links

• Website: https://rackmon.app
• Live demo: https://demo.rackmon.app
• Docs: https://rackmon.app/docs.html
• How to monitor rack temperature over SNMP: https://rackmon.app/guide.html
• Guía en español: https://rackmon.app/es/monitorizacion-temperatura-racks.html
• Compared with Zabbix / PRTG / LibreNMS / Checkmk: https://rackmon.app/compare.html
• Container image: https://hub.docker.com/r/rackmon/rackmon

Issues

Bug reports, deployment problems and OID corrections are all welcome here. For anything involving a licence key or account, email rackmon.software@gmail.com instead — please don't post keys in a public issue.

Found a security issue? See SECURITY.md — report it privately, not as an issue.