Automate remediation post AWS DevOps Agent investigation
aws ·
This article demonstrates how to automate incident remediation on AWS by combining the AWS DevOps Agent with Lambda Durable Functions, Amazon EventBridge, and Amazon Bedrock. While the DevOps Agent autonomously diagnoses root causes, organizations often keep it in an observe-and-report mode to prevent unintended changes. The proposed solution bridges this gap by triggering a durable function that analyzes investigation summaries and proposes specific fixes using a curated allowlist of approved Lambda tools. For read-only actions, the system executes them autonomously, but for infrastructure modifications, it suspends execution and waits for human approval. This approach significantly reduces mean time to resolution (MTTR) by presenting pre-validated changes ready for single-click approval. The workflow ensures safety through strict tool restrictions and maintains cost efficiency by pausing without consuming compute resources during approval delays. A practical example shows the system identifying a Lambda timeout issue and proposing a configuration update after verifying current settings.
aws