Agentic AI — Introducing the RCA Agent 'SentryOn' (Built with Strands SDK)

The Lotte On TechOps team introduced SentryOn, an RCA agent that autonomously reasons about the cause when a service failure occurs. The core idea is to use AI to automate the troubleshooting process in a complex MSA environment based on AWS EKS, which had previously relied on operators' experience and intuition.

In the existing incident response process, multiple people had to check different systems at the same time.

• DevOps engineers: AWS infrastructure and EKS
• DBAs: database status
• Developers: APM console and Elasticsearch logs

During business hours, multiple personnel can be deployed, but in situations such as late at night or when the responsible person is unavailable, response speed and accuracy inevitably suffer. Even with documented runbooks, there are limits to how quickly the actual bottleneck can be found among numerous services and infrastructure layers.

SentryOn was developed to shift this kind of repetitive incident analysis, which depends on specific personnel and experience, to Agentic AI. The article covers the process of introducing the RCA agent using Strands SDK and the technical challenges encountered while applying it in a production environment.