Автономне виявлення та усунення «сірих збоїв» у хмарних середовищах за допомогою агента ACSA

Abstract

This paper presents the Adaptive Control System Agent (ACSA) for the autonomous detection and targeted remediation of gray failures within cloud native environments. Traditional infrastructure monitoring often fails to detect such differential observability states, leading to undetected application degradation. To address this gap, the proposed architecture integrates continuous synthetic canaries with targeted configuration management. By leveraging serverless computing components, such as AWS Lambda, the agent executes rapid, isolated recovery actions without disrupting the underlying host instances. The implemented hierarchical escalation algorithm effectively isolates faults and reduces Mean Time to Recovery (MTTR) by over 98% compared to standard virtual machine reboots. Experimental validation confirms that this approach minimizes service degradation, ensures continuous high availability, and optimizes overall system reliability.