Network Inventory — Source of Truth on NetBox

Standing up NetBox is easy. Keeping it accurate over time — that's the work. This is the pipeline I use to bootstrap inventory and keep NetBox honest as the source of truth, without paying for a commercial sync tool.

The design has two engines: a Device Discovery Engine that finds nodes on the network, and a Device Configuration Check Engine that fingerprints them and writes them into NetBox.

Device Discovery Engine #

Walks each subnet and identifies live network gear:

1. Network scanning — sweep IP ranges to find responding hosts.
2. SNMP response check — for each live host, see if it answers SNMP.
3. sysDescription retrieval — pull MIB 1.3.6.1.2.1.1.1.0 to identify the platform.

import nmap
 
def scan_ports(ip):
    nm = nmap.PortScanner()
    nm.scan(ip, "22,161", arguments="-sU")
 
    for host in nm.all_hosts():
        print(f"Scanning {host}")
        for proto in nm[host].all_protocols():
            for port in nm[host][proto].keys():
                state = nm[host][proto][port]["state"]
                print(f"Port {port}/{proto} is {state}")
 
scan_ports("10.60.115.132")

$ snmpwalk -v3 -l authPriv -u SNMPV3User -a SHA -A ABCDEF -x AES -X ABCDEF \
    10.24.115.120 1.3.6.1.2.1.1.1.0
SNMPv2-MIB::sysDescr.0 = STRING: Cisco IOS Software [Cupertino], Catalyst L3 Switch Software (CAT9K_IOSXE), Version 17.9.4a, RELEASE SOFTWARE (fc3)

Python's subprocess library is perfectly fine if you want to shell out to snmpwalk instead of using a pure-Python SNMP client.

Device Configuration Check Engine #

Once a device's sysDescription tells us what platform we're on, the matching per-vendor check engine takes over:

1. Command execution — run a fixed set of show-commands to gather facts.
2. Payload preparation — translate the raw output into a NetBox-shaped object.
3. Device addition — POST it to the NetBox API.

Continuous synchronisation #

The whole loop runs every X hours/days, so NetBox keeps tracking real-world drift instead of slowly diverging.

Algorithm #

1. Start
2. Network scan (NMAP)
3. SNMP response check
   • If responsive: retrieve sysDescription → pick the right check engine → run commands → prepare payload → write to NetBox
   • If silent: skip to next IP
4. Wait X hours/days
5. Repeat

That's it — a small, boring pipeline that turns a green-field NetBox into a usable inventory and keeps it that way.