Automated Failure Detection and Container-Level Self-Healing for Dockerized Applications on AWS EC2 Using CloudWatch, Lambda, and Systems Manager

Container failures can interrupt services even when the underlying cloud virtual machine continues to operate normally. This study evaluates a lightweight self-healing approach for a Dockerized application deployed on Amazon EC2. The recovery workflow combines Amazon CloudWatch for failure detection, AWS Lambda for initiating remediation, and AWS Systems Manager for executing recovery actions on the EC2 instance. Three controlled failure conditions were evaluated: container crash, unhealthy application state, and frozen container. Each failure type was tested through ten manual and ten automated recovery trials, resulting in 60 formal trials. Recovery performance was evaluated using recovery time and successful restoration within a predefined 300-second threshold. All 30 automated trials recovered within the criterion, compared with 24 of the 30 manual trials. Automated recovery achieved a mean recovery time of 106.149 seconds compared with 288.298 seconds for manual recovery, representing a 63.18% reduction in mean recovery time. The results demonstrate that event-driven remediation can reduce operator-dependent recovery delays and provide more consistent recovery within the evaluated single-instance AWS environment.