Documentation › Integrations
Network devices over SNMP
An agent queries the devices of its network and publishes them as full hosts. SNMP secrets never leave the agent.
Declare a device
In the configuration of an agent that reaches the device, preferably a conf.d/20-snmp.yaml file (mode 600):
snmp:
enabled: true
interval: 60s
devices:
# SNMP v2c
- {name: rack-1-switch, address: 192.0.2.10, version: 2c, community: readonly, tags: {role: switch}}
# SNMP v3, authentication and encryption
- name: backup-nas
address: 192.0.2.20
version: 3
user: hypha
level: authPriv
auth_protocol: sha # md5, sha, sha224, sha256, sha384, sha512
auth_key: "…" # 8 characters at least
priv_protocol: aes # des, aes, aes192, aes256, aes192c, aes256c
priv_key: "…"
profile: auto
Other per-device settings: port (161), timeout (5 s), retries, context, interfaces_exclude (patterns), interface_down.
Advice:
- prefer SNMPv3 with
authPriv, with a read-only account limited to the agent's address; - disable the device's
publiccommunity before turning SNMP on; - check with
sudo hypha-agent check-config, then restart the agent.
Profiles
profile: auto reads sysObjectID and picks the profile:
| Profile | What it adds |
|---|---|
generic | MIB-II, IF-MIB, HOST-RESOURCES-MIB: interfaces, CPU, memory, storage |
mikrotik | RouterOS CPU, temperatures, fans, power supplies |
synology | disk, volume and RAID state |
| ASUSTOR (detected) | temperatures, fans, disk state and temperature, volume state and usage; snmp.disk.health and snmp.volume.health checks |
printer | printer state and supply levels |
ups | UPS-MIB: battery charge, runtime, output load |
Metrics and checks
| Metric | Unit |
|---|---|
snmp.ifBandwidthIn, .ifBandwidthOut | bytes per second |
snmp.ifBandwidthInUsage, .OutUsage | percent of the interface speed |
snmp.ifInErrors, .ifOutErrors, .ifInDiscards, .ifOutDiscards | packets per second |
snmp.ifOperStatus, snmp.ifAdminStatus | state |
snmp.cpu.usage, snmp.memory.usage, snmp.storage.usage | percent |
snmp.temperature, snmp.fan.speed, snmp.psu.status | degrees, RPM, state |
snmp.ups.battery.charge, .runtime, .load | percent, seconds |
snmp.sysUpTimeInstance, snmp.poll.duration | seconds, milliseconds |
Checks:
snmp.can_poll: critical when the device stops answering;snmp.interface.status: critical for an interface that is administratively up but down. Withinterface_down: seen(default), only an interface already seen up counts: a free switch port stays OK.alwaysapplies the rule to every up interface.
Traps
snmp:
traps: {enabled: true, listen: "0.0.0.0:162", community: readonly}
Each v1 or v2c trap received becomes an event (source:snmp type:snmp_trap), 60 per minute at most. An "event" monitor can alert on them.
In the interface
- Infrastructure filters by type (servers or network devices). A device page shows its identity and its interfaces as a table: name, alias, state, speed, throughput, errors, with small charts.
- Settings › Devices lists the devices seen, the agent that queries them and their last contact.
- The network map shows devices as nodes.
Catalogue templates (Network family): network interface down, device unreachable.
Updated: 19 November 2026