GitHub, a popular platform for developers to host and manage their code, experienced a major outage on August 17 that lasted nearly eight hours. During this time, users were unable to log in, use integrations, access APIs, or use GitHub Copilot, an AI-powered coding assistant. The outage was not caused by a software error or misconfiguration, but rather by an overwhelming surge in traffic that exceeded the system's capacity. The number of code commits—changes made to projects hosted on GitHub—has grown dramatically in recent months. In just four months, the monthly volume of commits nearly doubled, from 1.4 billion to 2.9 billion. As the demand for GitHub services increased, the main data center struggled to scale its infrastructure fast enough. When the system reached its peak load, its components failed to handle the demand, causing cascading failures. Specifically, Copilot clients unable to connect repeatedly tried to reconnect, creating a loop that overwhelmed the network and delayed the recovery of services. To address the issue, GitHub quickly deployed over three million additional processor cores and 120 petabytes of storage—equivalent to storing over 120 million books. The company also shifted a significant portion of its workload, moving 58 percent of its overall load and half of its Git operations to Microsoft’s Azure cloud platform. This migration, which had previously only involved 12 percent of GitHub’s operations, allowed the platform to adopt a more flexible architecture that can scale its read capacity in line with the number of users. In response to the outage, GitHub is overhauling its infrastructure to make it more resilient. The company plans to eliminate shared dependencies and isolate critical systems to prevent failures from spreading. New measures include implementing retry limits, request budgets, and variable expiration times for requests, all designed to prevent the kind of chain reactions that contributed to the outage. These changes aim to ensure that GitHub can handle future traffic surges without disrupting its users.