GitHub recently dropped their June 2026 Availability Report, and it’s a fascinating mix of wins, setbacks, and hard-earned lessons. If you’re into cloud infrastructure, DevOps, or just love a good behind-the-scenes look at how massive platforms keep the lights on, this one’s for you.
The Big Picture: Progress and Pauses
GitHub’s journey to improve their infrastructure reliability is a bit like climbing a mountain—it’s slow, deliberate, and sometimes you have to take a step back before moving forward. In June, they made some real structural progress, but they also hit a few speed bumps. Here’s what happened:
-
Monolith Traffic in Azure: GitHub’s goal is to shift more of their core services to Azure for better scalability and reliability. In June, monolith traffic in Azure peaked at 45% in the Central US region. While that’s not as high as they’d hoped, it’s still progress. The reason for the pause? A stability incident on May 21 forced them to hit the brakes and ensure the environment was stable before ramping up again. Smart move—better to go slow and steady than rush into a disaster.
-
Git in Azure: Git traffic (both HTTP and SSH) grew to 43%, but they missed their target of 50%. The reason? They’re prioritizing user experience by avoiding added latency. For example, they’re waiting on additional traffic routing and are currently only handling HTTP traffic for Git operations. It’s a reminder that sometimes you have to sacrifice speed for stability.
Under the Hood: Exciting Milestones
Despite the setbacks, there were some major wins worth celebrating:
- New Services: GitHub’s new extracted pull requests service, pullsd, is now handling 100% of anonymous pull request reads. That’s a big deal because it offloads traffic from the monolith, making the system more resilient.
- Reposd: This new repository service became the first to serve production REST traffic from Azure, ramping up to 50% of read traffic before being turned down due to a Redis capacity constraint. No rollbacks, no drama—just a controlled pause.
- API Rate Limiting: GitHub’s API rate limiting is now 97% handled at the Gateway, reducing contention with request-serving workers inside the monolith. That’s a huge improvement in efficiency.
- Security: Two-person confirmation is now required for all interactive production access and ChatOps changes, ensuring better oversight and accountability.
Incidents: The Other Half of the Story
No availability report is complete without a look at the incidents, and June had its share. GitHub experienced six incidents that degraded performance across various services. Here’s a quick rundown:
- June 4: Copilot code review failures due to an incompatible dependency. Lesson learned? Pin dependency versions and add better compatibility checks.
- June 8: Elevated HTTP 504 errors for signed-out users due to abusive automated traffic. GitHub improved automated detection and blocking to prevent future issues.
- June 10: API authentication failures caused by a memcached proxy misconfiguration. The fix? Migrate the authentication system to a more resilient caching infrastructure.
- June 16: Degraded availability for the Opus 4.8 model in Copilot due to an upstream provider issue. GitHub is now balancing capacity across providers to avoid single points of failure.
- June 17: Temporary unavailability of Copilot’s frontier chat models due to a faulty configuration change. The solution? Gradual rollouts with stronger validations.
- June 25: Background job service delays caused by hypervisor issues and traffic spikes. GitHub made the service more resilient and added earlier alerting to catch similar issues faster.
The Takeaway: Availability, Then Capacity, Then Features
GitHub’s guiding principle is clear: availability first, capacity second, features third. It’s a philosophy that prioritizes stability over flashy updates, and it’s paying off. By sharing these reports—warts and all—GitHub is setting a standard for transparency in the tech industry.
So, what’s next? More progress, more pauses, and more lessons learned. That’s the nature of the game. And as always, we’ll be watching and cheering them on.
Stay tuned for more updates, and let’s keep building a more reliable digital world together. 🚀