apim_throttle_expert

To diagnose 429 throttling errors, you need to correlate OperationId and ClientIpAddress in Log Analytics within a specific time window. This process identifies whether per-product quotas, per-key rate limits, or backend AOAI TPM exhaustion triggered the response.

A 429 throttling response occurs when a specific rate-limit policy is triggered. You can infer the root cause by examining throttling headers like x-throttling-source and x-aoai-throttle, which indicate whether the limit stems from product quotas or backend capacity.

Finding the root cause of Citadel API rate limiting involves mapping gateway logs to the operational trigger. By correlating the failing operation, client IP, and time window, you can classify the specific rate-limit source and return a recommended remediation action.

Yes, you can use Log Analytics to check backend AOAI TPM exhaustion during a 429 spike by correlating relevant gateway logs. This helps determine if the throttling originates from backend capacity limits rather than per-product or per-key rate-limit policies.

Diagnosing APIM throttling does not require modifying gateway policies or handling subscription keys. The diagnostic process is non-destructive, focusing on classifying the rate-limit source from gateway signals and recommending actions without applying changes.

The best way to classify APIM 429 errors for SRE on-call investigations is to analyze Retry-After and throttling headers. This approach pinpoints whether per-product quotas, per-key rate limits, or backend AOAI TPM exhaustion caused the failure, providing actionable output.