Bedrock InvokeAgent intermittent internalServerException (Sonnet 4.6, us-west-2)

Kindly reach out to the Bedrock AWS Support Team from the production account that you are getting the error. As mentioned by our re:Post Agent, AWS Support should be able to investigate the internal logs to determine the root cause and provide specific guidance for your queries.

The InternalServerException you're experiencing is a temporary condition caused by an unexpected internal error within the AWS service. According to AWS documentation, this error should resolve itself with retries, and implementing exponential backoff retry logic is the recommended approach—which you've already done at the Step Functions level.

Understanding the Error: When the Bedrock Agent Runtime service returns an InternalServerException (HTTP 500), it indicates a server-side issue rather than a problem with your configuration or request structure. The error can include a reason field; specifically, if the reason is BEDROCK_MODEL_INVOCATION_SERVICE_UNAVAILABLE, it means the model invocation service is temporarily unavailable and you should retry your request.

Regarding Your Specific Questions:

1. Capacity and Throttling: While the documentation confirms that InternalServerException is distinct from throttling exceptions (which return HTTP 429), I cannot provide real-time capacity information, SLOs, or confirm whether specific request IDs were throttled versus experiencing internal failures. These details require direct investigation by AWS Support with access to internal service logs.

2. Request-Specific Analysis: The specific failure reason for request ID be220290-ae2c-486d-b22d-6bd6a4c62fba would need to be investigated by AWS Support, as they have access to internal service telemetry that isn't exposed through public APIs or documentation.

3. Provisioned Throughput: I cannot confirm whether provisioned throughput would eliminate this specific error class without reviewing your exact use case with AWS Support. However, provisioned throughput generally provides more predictable performance characteristics.

4. Logging and Instrumentation: For the InvokeAgent API, you can enable trace information by setting enableTrace to true in your request. This will include agent reasoning processes, action group invocations, and knowledge base queries in the response. However, this won't expose the internal service failure reasons that cause the InternalServerException.

Recommended Next Steps: Given the production impact and the specific account/request IDs you've provided, I strongly recommend opening an AWS Support case with your current support plan. Include all the details you've shared here—request IDs, timestamps, account numbers, and the pattern analysis you've already completed. AWS Support can investigate the server-side logs to determine the root cause and provide specific guidance for your workload.
Sources
InternalServerException - botocore 1.42.50 documentation
InvokeAgent - Amazon Bedrock
InvokeAgentRuntimeCommandStreamOutput - Amazon Bedrock AgentCore Data Plane