SNMP Monitoring: Read Any Device by Its OID, Alert Before It Fails
SNMP monitoring reads a device's health — CPU, memory, disk, interfaces, uptime — straight from the agent it already runs, over one standard protocol. No per-metric scripts, no proprietary agents. Learn the OID tree once and it pays off across every server, switch, router and printer you own.
CPU
hrProcessorLoad
Disk used
hrStorageUsed
Uptime
hrSystemUptime
Now polling
1.3.6.1.2.1.25.3.3.1.2 → hrProcessorLoadExternal checks
SNMP monitoring at a glance
137
Reference pages
Protocol, sensors, setup, security, devices
3
SNMP versions
v1, v2c and v3 covered in depth
161
Default UDP port
snmpd listens here out of the box
40+
Documented OIDs
Real object identifiers with MIB names
Why SNMP Is the Default for Infrastructure Monitoring
SNMP wins because it is already there. The agent ships with, or installs in one line on, practically anything with an IP address — and it exposes standardized values any client can read. Here is what that buys you.
Agentless & already running
No bespoke collector per metric. Enable snmpd, point a poller at UDP 161, and you read CPU, memory, disk and interface counters through one common scheme that works the same on the next thousand boxes.
Pull-based & durable
Your monitor asks; the device answers. A healthy agent replies even while local tooling chokes — and when it goes silent, the silence itself is the alert. That is the case for watching from outside the network.
Standardized & portable
Every readable value has one canonical OID address defined by a MIB. A graph built for one Linux server works for the next. Learn the Host Resources and IF-MIB trees once, reuse them everywhere.
Explore the Knowledge Base
Ten pillars, each owning one slice of SNMP monitoring. Start with the protocol, wire up your agents, then make it durable by watching from outside.
How SNMP Monitoring Works
Four steps from a bare server to charted, alerting metrics. Every command below is valid Net-SNMP — the same syntax you will run in production.
Install the agent
The SNMP daemon is one package on most systems. It exposes the device’s values on UDP 161.
sudo apt install snmpd
sudo systemctl enable --now snmpdConfigure snmpd.conf
Pick a version, set a read-only community, and allowlist only the OIDs you actually poll.
# /etc/snmp/snmpd.conf
rocommunity s3cr3t-string 10.0.0.0/8
view all included .1.3.6.1.2.1.25Test from a client
Read a value by its OID before pointing a monitor at it. hrProcessorLoad returns per-core % load.
snmpget -v2c -c s3cr3t-string host \
1.3.6.1.2.1.25.3.3.1.2.1Poll, store & alert
Poll on a schedule, store history, set thresholds — ideally from outside the network for double durability.
# external checker polls every 60s
# silence from a dead agent = the alertWhat You Can Monitor via SNMP
CPU, memory, swap & load
hrProcessorLoad per core, the Host Resources storage table for RAM/swap, and laLoad from the UCD-SNMP-MIB — the vitals people watch every day.
Read the guideServer resources → charted & alerting
Disk space & capacity
Walk hrStorageDescr / hrStorageSize / hrStorageUsed to name each mount and chart usage — the metric that quietly kills web hosts.
Read the guideStorage → charted & alerting
Interface traffic & errors
64-bit ifHCInOctets counters from the IF-MIB for fast links, plus error and operational-status counters per interface.
Read the guideNetwork → charted & alerting
Temperature, power & UPS
Entity Sensor MIB readings for temperature, fans, voltage and PDU/UPS state on enterprise hardware.
Read the guideEnvironment → charted & alerting
Polling vs Traps vs External Monitoring
SNMP data can reach your monitor three ways, and where the monitor lives matters as much as how it collects. Most resilient setups combine them.
| Criterion | Internal polling | SNMP traps | External pollingRecommended |
|---|---|---|---|
| Direction | Manager pulls | Agent pushes | Manager pulls |
| Sees a dead box | No | No (can’t send) | Yes — silence is the alert |
| Setup effort | Low | Medium | Zero (SaaS) |
| Latency to alert | Poll interval | Instant | Poll interval |
| Best for | Dashboards | State changes | Availability + durability |
External polling is highlighted because a monitor inside the failure domain can’t warn you when that domain fails — the core idea behind double durability.
Frequently Asked Questions About SNMP Monitoring
Do I need to install software on every server?
Only the SNMP agent, which most systems already have or can add in one command. There is no per-metric script and nothing proprietary — the manager side just reads the values the agent already publishes.
Which SNMP version should I run?
SNMPv2c is the common choice for read-only polling on a trusted network. Use SNMPv3 when you need authentication and encryption on the wire. See the versions breakdown for the trade-offs.
Why poll from outside my own network?
A monitor inside the failure domain can’t warn you when that domain fails. An external checker still gets an answer from a healthy agent and treats silence from a dead one as the alert — that is double durability.
Learn the protocol, wire up a hardened agent, then make monitoring durable by watching from outside the network with ostr. Explore the knowledge base.
Start reading a device by its OID
Learn the protocol, wire up a hardened agent, then make monitoring durable by watching from outside the network with ostr.io.