We will examine the thought process and workflow I followed while investigating an outage that impacted hundreds of thousands of users. Beneath the flood of 500-errors was concealed a deeper problem: Redis logs reported closed sockets, VPC Flow Logs displayed thousands of connections, and CloudWatch showed a saturated NAT Gateway. The breakthrough came from an AI‑chat insight about a recent third‑party failover‑strategy change, which pointed directly to a small code bug that triggered a massive connection storm.

In this session you will learn how to treat metrics and logs as evidence, filter out noise, and adopt a detective’s mindset to uncover the true root cause. You will leave with practical techniques to reconstruct incident timelines, leverage AI for hidden insights, apply investigative methods to complex infrastructure failures, and, ultimately, produce an effective RCA.