Browsing is open · Catalog captured 2026-11-21 · Automatic updates and account services are paused.
Design & UI
netbox-review-datamodel
Audit and review NetBox data model design choices. Use when evaluating site hierarchy, IPAM organization, device modeling, tenancy patterns, custom field vs tag decisions, naming conventions, and extensibility strategy. Identifies modeling anti-patterns and suggests improvements aligned with NetBox best practices.
name: netbox-review-datamodel
description: >
Audit and review NetBox data model design choices. Use when evaluating site hierarchy,
IPAM organization, device modeling, tenancy patterns, custom field vs tag decisions,
naming conventions, and extensibility strategy. Identifies modeling anti-patterns and
suggests improvements aligned with NetBox best practices.
license: Apache-2.0
NetBox Data Model Review
Your knowledge of NetBox data models may be outdated. Available model types, relationships, and extensibility mechanisms evolve between releases. Verify against current docs before recommending changes.
If configured — inspect the current data model, count objects, find inconsistencies
Live model audit
Review Workflow
Follow this checklist when auditing a NetBox data model:
Understand the organization — what kind of network? (enterprise, DC, ISP, hybrid)
Check hierarchy — is the Region → Site Group → Site → Location → Rack hierarchy appropriate?
Check IPAM — are prefixes properly nested? VRFs used where needed? Aggregates defined?
Check naming — are slugs consistent? Are names human-readable and grep-friendly?
Check extensibility choices — custom fields vs tags vs config contexts vs custom objects
Check tenancy — is multi-tenancy used correctly? Over-tenanted or under-tenanted?
Check for modeling smell — overloaded fields, misuse of description for structured data
Critical Rules
Hierarchy (HIER)
ID
Rule
Severity
HIER-1
Use Regions for geography, Site Groups for function — don't conflate
High
HIER-2
Locations are recursive — don't flatten (Floor > Room > Row is better than Row alone)
Medium
HIER-3
Every device should be in a Site (never orphaned at top level)
High
HIER-4
Racks should have a Location when the site has locations defined
Medium
HIER-5
Don't create single-child hierarchies — they add complexity without value
Low
HIER-6
On 4.6+, RackGroups offer a flat, cross-location way to group racks (e.g. "cage-7", "cold-aisle-B") independent of the Location tree — use them when rack grouping doesn't map cleanly onto site/location nesting, not as a substitute for Locations
Low
HIER-7
On 4.7+, hierarchies are ltree-backed: flag scripts/saved filters/exports that filter or order by level (no longer a DB column) — use parent_id/ancestor_id or get_ancestors() instead. Modeling advice is unchanged
Medium
HIER-8
On 4.7+, if the instance was restored from a 4.7.0pg_dump, descendant paths may be stale after renames/moves — have the admin run manage.py rebuild_ltree_paths --check and repair before trusting hierarchy reports
Medium
IPAM (IPAM)
ID
Rule
Severity
IPAM-1
Define RIR Aggregates to track address space allocation boundaries
Medium
IPAM-2
Prefixes should nest correctly (child within parent CIDR)
High
IPAM-3
Use VRFs to separate overlapping address spaces — never duplicate prefixes in global table
Critical
IPAM-4
IP Addresses must have prefix length (/32 for loopbacks, actual mask for interfaces)
High
IPAM-5
Use Roles to classify prefix purpose (infrastructure, customer, management); on NetBox 4.6+ ASNs can also carry a Role — use it to classify ASN purpose
Medium
IPAM-6
VLANs should be in VLAN Groups with an appropriate scope. Scope can be Region/SiteGroup/Site/Location/Rack/ClusterGroup/Cluster (and RackGroup on 4.6+) — don't assume site/location are the only options; pick the tightest scope that matches the VLANs' reuse boundary
Medium
IPAM-7
On 4.7+, a service that listens on one port over several protocols (DNS tcp/53 + udp/53) is one Service with multiple port_mappings — flag duplicate per-protocol services (dns-tcp, dns-udp) and merge them; flag integrations still writing the deprecated protocol/ports pair (removed 5.0). On 4.5–4.6 one-service-per-protocol is the only option
Low
Device Modeling (DEV)
ID
Rule
Severity
DEV-1
Every device needs a Device Type (not just a name) — this enables port planning
High
DEV-2
Use Roles to classify function (router, switch, firewall) — not naming convention alone
High
DEV-3
Platforms indicate software — assign them for config template compatibility
Medium
DEV-4
Virtual chassis members should have proper VC position and master assignment
Medium
DEV-5
Interface types must match reality (1000base-t vs 10gbase-sr) for capacity planning
Medium
DEV-6
Every Rack should have a Rack Type with dimensions defined there. On 4.7 the per-rack form_factor, width, and outer_* fields are deprecated; in 5.0 they are removed and rack_type becomes mandatory — flag racks whose rack_type is null (walk GET /api/dcim/racks/?brief=1 — there is no null filter on rack_type_id) and per-rack dimensions that differ from their type
High
DEV-7
On 4.7+, model breakout/channelized ports with channels on the parent and channel_id subinterfaces (channel type), not as loose virtual interfaces or extra physical interfaces — otherwise cable paths and port counts are wrong
Medium
DEV-8
On 4.7+, use Module Bay Types where a chassis restricts which modules fit which slots (line-card vs supervisor vs fan bays) — an empty bay-type set means "anything fits", so absence is fine for simple hardware but flag cha