The new traffic spike caused “capacity outages” which led to the recent dev platform outage.
Following its extended outage on Monday, GitHub says capacity issues – not a code or configuration change – led to many of its services being inaccessible for more than seven hours. In the developer platform’s own words, “traffic reached a new high and a critical infrastructure component in our central US data center failed to scale with it.”
The change in demand that has caused the increase in traffic is dramatic. “Since April, monthly commitments have increased from 1.4 billion to 2.9 billion,” says GitHub. “This growth explains the pressure on our systems, but it does not excuse these outages.” This means that the number of changes made to projects hosted on the platform has more than doubled in the last four months, a huge increase in usage that tracks the growing amount of code produced by AI.
GitHub offers a more technical explanation of what went wrong on its status reporting page, but the most important fixes the platform has made to prevent future outages are in how it handles concurrent activities. Specifically, GitHub says it examines CPU and memory alerts to detect components of its servers that may fail during periods of high traffic, and that it “enforces consistent retry limits, retry budgets, and variable timeouts in service-to-service interactions to avoid retry storms and cascading loads.”
While these address GitHub’s immediate problems, they don’t address the trend that caused the traffic spike in the first place: more and more code is being pushed to AI. As Anthropic and OpenAI continue to build on their AI coding tools, the pressure identified by GitHub is unlikely to go away. As the default site for professional and amateur software projects, GitHub is able to continue to feel it.
