Source Of Truth Audit

Source-of-Truth Reconciliation Audit

Network source-of-truth (SOT) reconciliation audit for comparing intended infrastructure state recorded in [NetBox] or [Nautobot] against live network reality. This skill provides a systematic reconciliation methodology — what to extract, what to compare, how to classify gaps, and how to score data quality — not a comprehensive API tutorial for either platform.

SOT platforms maintain the authoritative record of what the network should look like: device inventory, interface assignments, IP allocations, VLAN mappings, site/rack topology, and cable records. Live network state — collected via SNMP, CLI, or streaming telemetry — represents what the network actually looks like. The gap between intent and reality is the reconciliation target. Where API patterns diverge between platforms, [NetBox] and [Nautobot] inline labels mark the platform-specific approach while shared logic remains unlabeled.

Reference references/cli-reference.md for [NetBox] REST API and pynetbox patterns, [Nautobot] REST API and pynautobot patterns, plus SNMP and CLI discovery commands for live state collection. Reference references/reconciliation-workflow.md for the diff methodology, gap classification taxonomy, and data quality scoring framework.

When to Use

• Periodic SOT hygiene audit — scheduled reconciliation to measure and improve data quality
• Pre-automation gate — validating SOT accuracy before trusting it as the source for automation pipelines
• Post-change verification — confirming that executed changes are reflected in the SOT within the expected sync window
• Network migration planning — establishing a baseline of what the SOT knows vs what exists before migration
• Incident root-cause analysis — determining whether a SOT discrepancy contributed to an outage
• Compliance audit preparation — demonstrating that the network inventory is accurate and complete for audit requirements

Prerequisites

• [NetBox] access — API token with read permissions on DCIM, IPAM, circuits, and tenancy modules; base URL confirmed (curl -s -H "Authorization: Token <token>" https://<netbox>/api/status/ returns version)
• [Nautobot] access — API token with read permissions on DCIM, IPAM, circuits, and tenancy; base URL confirmed (curl -s -H "Authorization: Token <token>" https://<nautobot>/api/status/ returns version); GraphQL endpoint at /api/graphql/ for bulk extraction
• Live network discovery credentials — SNMP v2c community strings or v3 credentials for device polling; SSH/NETCONF credentials for CLI collection; read-only access sufficient
• Device scope defined — target sites, device roles, and platforms identified before extraction; large environments should scope by site or region
• Baseline expectations — understand the SOT's intended coverage: does it track all devices or only core/distribution infrastructure? Findings are only meaningful against defined coverage scope

Procedure

Follow these six steps sequentially. Each step builds on the previous — SOT extraction and live discovery feed the diff analysis, which feeds gap classification and quality scoring, culminating in the reconciliation report.

Step 1: SOT Inventory Extraction

Extract the authoritative device inventory, interface assignments, IP allocations, and topology data from the SOT platform. The goal is a complete snapshot of intended state for the in-scope devices.

[NetBox] Use the REST API to extract device records with all relevant attributes. Filter by site or region to scope the extraction:

GET /api/dcim/devices/?site=<site-slug>&limit=1000
GET /api/dcim/interfaces/?device_id=<id>&limit=500
GET /api/ipam/ip-addresses/?device_id=<id>
GET /api/ipam/prefixes/?site=<site-slug>
GET /api/dcim/cables/?device_id=<id>

Use ?include=custom_fields to capture organization-specific metadata. Paginate using offset and limit parameters — check the count field in responses to ensure complete extraction. For large sites, pynetbox bulk retrieval is more efficient than raw curl pagination (see references/cli-reference.md).

[Nautobot] Use the REST API for the same data extraction. Nautobot endpoint structure mirrors NetBox with minor namespace differences:

GET /api/dcim/devices/?location=<location-slug>&limit=1000
GET /api/dcim/interfaces/?device=<uuid>
GET /api/ipam/ip-addresses/?device=<uuid>
GET /api/ipam/prefixes/?location=<location-slug>

For efficient bulk extraction, [Nautobot] offers a GraphQL endpoint that can retrieve devices with nested interfaces, IPs, and cables in a single query — significantly reducing API call volume for large inventories. See references/cli-reference.md for the GraphQL query pattern.

Record the extraction timestamp — this becomes the SOT snapshot time for freshness calculations in Step 5.

Step 2: Live Network Discovery

SNMP Polling — Walk sysName, sysDescr, sysLocation, and ifTable/ifXTable for each device in the target subnet range. sysName.0 provides the hostname for entity matching against SOT records. ifAdminStatus and ifOperStatus per interface provide admin/operational state for comparison against SOT interface status.

CDP/LLDP Neighbor Discovery — Collect neighbor tables to map physical connectivity. Compare against SOT cable records to identify undocumented connections or missing cable entries. On Cisco devices: show cdp neighbors detail; on multi-vendor environments: show lldp neighbors detail.

IP Address Collection — Enumerate configured IP addresses per interface: show ip interface brief (Cisco), show interfaces terse (Juniper). Compare against SOT IP address assignments.

ARP/MAC Table Collection — Capture ARP and MAC address tables to detect endpoint presence and validate SOT-documented connections: show ip arp, show mac address-table.

Routing Table Snapshot — Collect routing tables to verify prefix advertisements match SOT prefix allocations: show ip route summary, show ip bgp summary.

Step 3: Diff Analysis

Compare SOT-extracted records against live discovery results across five entity types, each with its own matching criteria and comparison attributes.

Device Existence — Match SOT device records against live SNMP sysName responses. Classify results as:

• Matched: Device exists in both SOT and live discovery
• SOT-only: Device in SOT but no live response (potentially decommissioned)
• Live-only: Device responding but not in SOT (undocumented/shadow IT)

Entity matching uses hostname as the primary key. When hostnames are ambiguous, fall back to serial number or management IP as secondary match keys. See references/reconciliation-workflow.md for the matching algorithm.

Interface State — For matched devices, compare interface-level attributes: SOT admin status vs live ifAdminStatus, SOT-documented description vs live ifAlias, SOT speed/duplex vs live ifHighSpeed. Flag interfaces where SOT shows "active" but live shows ifOperStatus down — these are the highest-priority mismatches for operational impact.

IP Address Assignment — Compare SOT IP allocations against live interface IP addresses. Flag addresses configured on devices but not recorded in the SOT (undocumented allocations), and SOT allocations with status "Active" but no corresponding live configuration (phantom allocations).

VLAN Assignment — Compare SOT VLAN-to-interface mappings against live switchport configurations. Identify trunk ports carrying VLANs not documented in SOT and access ports assigned to VLANs different from SOT records.

Topology and Connectivity — Compare SOT cable records against CDP/LLDP neighbor data. Identify physical connections observed by LLDP but absent from SOT cable records (undocumented cabling), and SOT cable records with no matching LLDP neighbor (stale or incorrect cable documentation).

Step 4: Gap Classification

Classify diff results into actionable categories with different operational implications and remediation approaches.

SOT-Only Entries (Stale Records) — Devices, interfaces, or IPs in the SOT but not in live discovery. Common causes: decommissioned equipment not cleaned from SOT, planned devices not yet deployed, or discovery scope not reaching the device. Validate with targeted ping or console access before classifying as truly stale.

Live-Only Entries (Undocumented Assets) — Devices, interfaces, or IPs discovered live but absent from the SOT. Common causes: emergency deployments not entered into SOT, lab equipment, shadow IT, or devices outside the SOT's intended scope. These represent the highest SOT integrity risk — undocumented devices cannot be managed or automated.

Attribute Mismatches — Matched entities with conflicting attribute values: wrong serial number, incorrect model, outdated firmware version, wrong site/rack/position, or stale interface descriptions. Mismatches degrade SOT trustworthiness even when the entity is documented.

Relationship Gaps — Missing or incorrect relationships between entities: cable records absent for LLDP-confirmed connections, prefix assignments missing VLAN associations, circuits not linked to interfaces. Relationship gaps break topology visualization and impact-analysis accuracy.

See references/reconciliation-workflow.md for the full gap classification taxonomy with severity mappings.

Step 5: Data Quality Scoring

Compute quantitative data quality metrics across four dimensions to track SOT health over time and identify which areas need the most remediation attention.

Completeness — Percentage of required fields populated per device record. Required fields vary by device role: core routers need BGP ASN and loopback IPs; access switches need port profiles and VLAN assignments. Calculate as populated_required_fields / total_required_fields × 100% per device, then average across the inventory.

Accuracy — Percentage of populated SOT fields that match live state. Only measurable for attributes observable via discovery (hostname, model, serial, interface state, IP assignment). Calculate as matching_fields / discoverable_populated_fields × 100%. Accuracy below 85% indicates systematic SOT update failures.

Freshness — Days since last_updated timestamp per record. [NetBox] exposes last_updated on all objects via the API. [Nautobot] provides last_updated with the same semantics. Flag records not updated in 90+ days as potentially stale — especially for dynamic attributes like interface operational state.

Coverage — Percentage of live-discovered devices with corresponding SOT records. Calculate as matched_devices / total_live_devices × 100%. Coverage below 95% means the SOT has blind spots for automation relying on its inventory.

Aggregate the four dimensions into a composite SOT Health Score for trending. See references/reconciliation-workflow.md for the scoring framework with threshold recommendations.

Step 6: Reconciliation Report

Compile findings into a structured report with executive summary, gap inventory, data quality scorecard, and prioritized remediation.

Executive Summary — Total devices in scope, match rate, gap counts by category (SOT-only, live-only, mismatch, relationship), composite data quality score, and comparison to prior reconciliation if available.

Gap Inventory — Complete list of gaps organized by classification (Step 4). Each entry includes the affected entity, gap type, severity, and recommended remediation. Prioritize live-only devices and interface state mismatches at the top.

Data Quality Scorecard — Present the four quality dimensions as percentages with trend arrows if prior data exists. Break down scores by site or device role to identify which segments drag down overall quality.

Remediation Priorities — Rank gaps by operational impact: (1) undocumented live devices → create SOT records immediately, (2) decommissioned device records → archive or delete, (3) attribute mismatches → update SOT from live data, (4) relationship gaps → rebuild cable/circuit records from LLDP. Estimate effort per category to enable sprint planning.

Threshold Tables

Decision Trees

SOT Platform Selection for Extraction:
├─ [NetBox] deployed? → Use REST API with pagination
│  ├─ Large inventory (>5000 devices)? → Use pynetbox bulk retrieval
│  └─ Custom fields in use? → Include ?include=custom_fields
├─ [Nautobot] deployed? → Choose extraction method:
│  ├─ Need nested relationships? → Use GraphQL endpoint
│  └─ Simple entity lists? → Use REST API
└─ Both deployed? → Reconcile between platforms first, then vs live

Gap Severity Assignment:
├─ Live-only device (undocumented)?
│  ├─ Production network? → Critical — immediate SOT update required
│  └─ Lab/staging? → Medium — document within standard cycle
├─ SOT-only device (stale)?
│  ├─ Shows as active/planned status? → High — misleads automation
│  └─ Shows as decommissioned? → Low — cleanup task
├─ Attribute mismatch?
│  ├─ Hostname or management IP? → High — breaks automation lookups
│  ├─ Serial/model/firmware? → Medium — asset tracking impact
│  └─ Description or rack position? → Low — cosmetic
└─ Relationship gap?
   ├─ Missing cable record for active link? → Medium — topology blind spot
   └─ Stale cable record for removed link? → Low — cleanup task

Discovery Method Selection:
├─ Managed devices with SNMP? → SNMP polling (fastest, most complete)
├─ SNMP blocked by policy? → CLI collection via SSH/NETCONF
├─ Need physical topology? → CDP/LLDP neighbor tables (required)
├─ Need endpoint presence? → ARP/MAC table collection
└─ Need routing state? → Route table + BGP summary

Report Template

# Source-of-Truth Reconciliation Report

## Executive Summary
- **SOT platform:** [NetBox/Nautobot] version [X.Y]
- **Scope:** [sites/regions] — [device count] devices in scope
- **Reconciliation date:** [date]
- **Composite SOT Health Score:** [score]%

## Data Quality Scorecard
| Dimension    | Score | Trend | Threshold |
|-------------|-------|-------|-----------|
| Coverage     | [%]   | [↑↓→] | ≥95% good |
| Accuracy     | [%]   | [↑↓→] | ≥90% good |
| Completeness | [%]   | [↑↓→] | ≥85% good |
| Freshness    | [days] | [↑↓→] | <30d good |

## Gap Summary
| Category | Count | Critical | High | Medium | Low |
|----------|-------|----------|------|--------|-----|
| SOT-only (stale) | | | | | |
| Live-only (undoc) | | | | | |
| Attribute mismatch | | | | | |
| Relationship gap | | | | | |

## Gap Inventory

### Critical — Undocumented Live Devices
| # | Hostname | mgmt IP | Discovered Via | Site | Action |
|---|----------|---------|----------------|------|--------|

### High — Stale Active Records
| # | SOT Device | Status in SOT | Last Ping | Action |
|---|------------|---------------|-----------|--------|

### Medium / Low — Attribute & Relationship Gaps
[Grouped table with entity, gap type, current SOT value, live value]

## Remediation Priorities
1. [Immediate — Create SOT records for undocumented devices]
2. [Short-term — Archive stale device records]
3. [Medium-term — Resolve attribute mismatches]
4. [Ongoing — Establish automated sync or webhook triggers]

## Appendix
- Extraction parameters and API query details
- Discovery methodology and SNMP/CLI command inventory
- Scoring framework and threshold definitions

Troubleshooting

SOT API rate limiting — [NetBox] and [Nautobot] may throttle API requests under heavy extraction. Both return HTTP 429; implement backoff in extraction scripts. Use pagination with modest page sizes (100–500) rather than requesting all records in a single call. GraphQL on [Nautobot] reduces total API call volume.

SNMP discovery returns partial inventory — Ensure SNMP credentials match across all device platforms. Devices behind firewalls may need SNMP access rules. Verify reachability with snmpget -v2c -c <community> <ip> sysName.0 before full walks.

Entity matching failures on hostname — Matching breaks when devices have inconsistent naming between SOT and live SNMP sysName. Common causes: FQDN in SOT but short name in SNMP, case mismatches, or domain suffix differences. Normalize both sides (lowercase, strip domain) before comparison. Fall back to management IP or serial number for unresolved entries.

Large environment performance — Reconciliation of 10,000+ devices generates significant data volume. Batch by site or region. Run SOT extraction and live discovery in parallel to minimize the snapshot window — a 4-hour gap introduces drift that appears as false mismatches.

Freshness scores skewed by bulk imports — If the SOT was recently bulk-imported, all last_updated timestamps reflect the import date. Flag this in the report and exclude freshness from the composite score until one reconciliation cycle updates records individually.

[NetBox] vs [Nautobot] field name differences — Key divergences: [NetBox] uses site while [Nautobot] uses location; [NetBox] uses integer IDs while [Nautobot] uses UUIDs. Normalize field names before feeding into the diff engine. See references/cli-reference.md for the field mapping table.