GitHub admits April outages, AI-powered workflows pushed demand 30 times
GitHub had a rough April after AI-powered workflows pushed its systems way past their limits: demand spiked to 30 times what was expected, and uptime dropped below 85% (not great when you promise 99.9%).
Chief Technology Officer Vlad Fedorov owned up to the mess, admitting core services like Git storage just weren't built for this kind of surge.
GitHub upgrades systems after April incidents
Two big incidents made things worse: a bug on April 23 caused mix-ups in more than 2,000 pull requests, and then on April 27, a possible botnet attack overloaded search features.
To fix things, GitHub is upgrading its systems with better caching and service isolation, moving some code to Go for speed, and updating the status page to include availability numbers and more transparent incident reporting.
The goal? Make sure this kind of chaos doesn't happen again as AI keeps growing.