SNMP and the NMS: monitoring that tells the truth
▶ Start the interactive lesson
Friday 16:50. The dashboard is all green. The phones are ringing anyway: the intranet is down. Maya scrolls the monitoring config: "We ping the server. It answers. Ping can't tell you the web service died, or that a link is at 99% with errors. Let's monitor what users actually feel."
Key points
- SNMP: the NMS polls devices on UDP 161 (interface status, traffic, errors, CPU, memory) and devices send traps to the NMS on UDP 162 the moment something happens (link down, reboot). Data lives in the MIB, addressed by OIDs (e.g. ifOperStatus).
- Versions: v1/v2c use a plain-text community string as a password (read-only RO or read-write RW). v3 adds users with authentication and encryption (authPriv). Use v3 where you can. With v2c: RO only, never "public"/"private", and lock it to the NMS with an ACL.
- What to monitor: availability (ping) is the bare minimum. Add service checks (HTTPS returns 200, DNS resolves the name, a TCP port answers), interface status, utilisation and errors, CPU/memory, and environment (PSU, fans, temperature).
- Monitoring traffic crosses the network like anything else: routes, ACLs, firewall policies and VPN selectors must allow it, ideally from the NMS's IP only. A new monitoring subnet is a classic thing to forget in a VPN.
- Alerting: set thresholds against a baseline, alert the right team, and avoid alert fatigue (if everything alerts, nothing does). Popular NMS tools: PRTG, LibreNMS, Zabbix, SolarWinds, Auvik and LogicMonitor.
The full lesson has more: animations, a hands-on exercise, 3 quiz questions and spaced review cards. Open it free →