Zabbix monitors an industrial switch by polling it over SNMPv3 every 30–60 seconds and receiving asynchronous traps on UDP 162, then graphing port traffic, CPU, temperature, and — on SFP-equipped models — digital diagnostic (DDM) optical power. Rayin industrial switches speak standard MIB-II / IF-MIB and SNMPv3, so they drop straight into an existing Zabbix instance without custom agents — the same SNMP interface also covers Rayin mini OLTs, so a campus network is monitored from access switch to OLT on one screen.
Scale: beyond ~10 devices, manual ping checks do not scale; SNMP polls automatically and draws trends.
Real-time: polling is pull, but a trap (UDP 162) is push — link down or over-temperature is reported the moment it happens.
Visibility: SFP DDM reads Rx optical power and temperature, catching a failing fiber patch before users notice.
SNMP has three versions; v1/v2c send the community string in clear text, while v3 (RFC 3414 USM) adds SHA authentication and AES encryption — mandatory on an industrial network.
Use SNMPv3 (RFC 3414) with AES; never leave v1/v2c open.
Poll every 30–60s and enable traps (UDP 162) for instant fault push.
Watch DDM Rx power: alarm below −28 dBm, overload above −3 dBm.
Enable the SNMP agent on UDP 161.
Create an SNMPv3 user with SHA authentication and AES encryption (DES optional); disable v1/v2c.
Limit the source IP to the NMS segment (e.g. 192.168.99.0/24).
Move Web/SSH/SNMP onto the dedicated management VLAN 99, physically separate from business domains.
Rayin industrial switches expose standard MIB-II, so Zabbix can read them through the built-in template set.
Create a Host; under Interfaces choose SNMP and enter the switch management IP and port 161.
Set SNMP version to v3, then enter the username, auth/encrypt algorithms, and keys — the context and engine ID must match the switch exactly.
Verify with snmpwalk against sysDescr (MIB-II RFC 1213, OID 1.3.6.1.2.1.1.1) before linking templates.
IF-MIB: per-port bytes in/out, errors, discards, utilization — auto-graphed.
CPU / Memory: alert at CPU >80% and memory >85%.
SFP DDM: read Rx power; normal window −3 to −28 dBm, alarm below −28 dBm (approaching receive sensitivity).
LLDP (IEEE 802.1AB): auto-discovers neighbors and draws topology.
Point Zabbix at an snmptrapd receiver so a link OperStatus down alerts immediately.
Key thresholds: port packet loss >1%, temperature >70°C, optical Rx <−28 dBm or >−3 dBm (overload).
Route alerts to email / DingTalk / Webhook so the on-call engineer is paged, not the morning report.
Confirm Latest Data refreshes continuously (polling works).
Pull a week of traffic to find busy-hour peaks and plan capacity.
Shut one port and confirm a trap fires within a minute.
Templates: package items, triggers, and graphs; assign to a host and it is monitored. Official Template Net covers common MIBs; Rayin switches map cleanly to MIB-II / IF-MIB.
Low-Level Discovery (LLD): auto-create items for every up port and SFP via IF-MIB ifDescr/ifIndex; new line cards need no manual registration.
Trigger functions beyond thresholds: last(), avg(5m), min(10m)/max(10m), count(10m,"down"), and forecast(1h)/timeleft() — the last predicts when Rx power will cross −28 dBm so you swap the patch first.
Value mapping & user macros: map IF-MIB OperStatus 1/2/3 to up/down/testing; set {$IF.UTIL.MAX}=80, {$TEMP.MAX}=70 so one template fits every workshop by changing macros only.
Dashboard & Maps: LLDP auto-draws "Switch A ↔ OLT B" links; a red link locates the break faster than log digging.
Zabbix Proxy: a remote factory runs a Proxy in active mode, caching and batching back to the HQ server — bandwidth-friendly and outage-safe.
Autoregistration + API: new switches with SNMPv3 auto-classify and link templates by IP/LLDP; the Zabbix API (JSON-RPC) syncs CMDB changes with zero manual entry.
Maintenance & Escalation: silence alerts during planned upgrades; escalate unacknowledged faults from DingTalk to email to phone.
Typically 30–60 seconds. Core devices can go to 10–15s, but do not over-poll or you load the agent's CPU.
Yes. Polling shows trends; traps catch sudden changes. They complement each other.
Some older transceivers lack DDM; swap in a DDM-capable SFP. Most modern managed switches expose DDM on their SFP ports, so the reading appears once the IF-MIB DDM template is linked.
Authentication fails and Zabbix reports timeouts. The engine ID must match the switch configuration exactly.
Use templates + Low-Level Discovery + Autoregistration; new devices self-register, and the Zabbix API can sync CMDB entries automatically.
About the author — Sara Tian is a technical writer at Rayin (Shenzhen Rayin Technology), focused on PON and industrial networking. Connect with Sara on LinkedIn.
About Rayin → Rayin company profile
