Error Handling

The Data Collector never aborts. Every API call failure is caught, classified, and written to the collection_errors array in the output JSON. Collection continues regardless of individual failures.

Error Classification

HTTP Status	Classification	Collection behaviour	Assessment effect
403 Forbidden	Permission missing	Log error, continue	Affected items → data_unavailable
404 Not Found	Resource not deployed	Log as not_applicable, continue	Affected items → not_applicable
429 Too Many Requests	API throttling	Exponential backoff, 3 retries, then log	Affected items → data_unavailable if retries exhausted
408 / Timeout	Network timeout	30s timeout per call, log, continue	Affected items → data_unavailable
500 / 503	Server error	Single retry after 5s, then log	Affected items → data_unavailable

Retry Policy

• Maximum retries per call: 3
• Backoff strategy: exponential with jitter — base 1s, max 16s
• Retry applies to: 429, 500, 503, timeout
• No retry on 403 or 404 — these are deterministic

collection_errors Schema

{
  "phase":           "rest",
  "call":            "GET /subscriptions/{id}/providers/Microsoft.Security/assessments",
  "subscription_id": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "http_status":     403,
  "error_code":      "AuthorizationFailed",
  "message":         "The client does not have authorization...",
  "timestamp":       "2026-03-17T14:23:01Z",
  "retry_count":     0
}