GitHub is facing the challenge of massive growth as the proliferation of artificial intelligence and autonomous agents has overwhelmed its infrastructure, leading to a nearly eight-hour outage on August 17, its second major incident in a month. In response, the company is accelerating the migration of its services to the Microsoft Azure cloud.
GitHub is facing the challenge of massive growth as the proliferation of artificial intelligence and autonomous agents has overwhelmed its infrastructure, leading to a nearly eight-hour outage on August 17, its second major incident in a month. In response, the company is accelerating the migration of its services to the Microsoft Azure cloud.
According to new data, GitHub's monthly workload has reached impressive proportions: the platform is processing 2.9 billion commits, which is more than double the April figure of 1.4 billion, 130 million merged pull requests, and 24 million new repositories, writes The New Stack.
The reason for this explosive growth is the rapid development of AI-based development tools and a change in the very approach to software development. Despite efforts to transition to a more modern architecture, the platform is unable to scale quickly enough.
According to a technical analysis from GitHub CTO Vlad Fedorov, the failure was caused purely by scaling issues, not by code changes.
“Our investigation showed that the outage began when traffic reached a new peak, and a critical infrastructure component in the Central US data center was unable to scale with it. As a result, the lack of capacity spread throughout all systems, causing authentication errors and failures in the operation of many GitHub services,” Fedorov explained.
The problem wasn't due to a lack of resource expansion. According to the CTO, the company has already moved 58% of the platform's workload to Azure (up from just 12% in May), and has also moved half of all Git operations to the cloud.
This year, GitHub has added over 3 million processor cores and 120 petabytes of high-speed storage.
However, the service's own data centers have reached their limit: as much equipment has been installed as the available energy capacity allows.
Fedorov also acknowledged that governance processes have not kept pace with the pace of change: the team has already redirected resources to improve availability, strengthen testing, monitoring, and notifications. In addition, work is ongoing to isolate critical systems and remove shared dependencies to reduce the likelihood of cascading failures.
To prevent "request storms" and infrastructure overload, GitHub already implements hard limits on connection retries and adjusts timeouts between services.
