GitHub availability report: May 2026
GitHub says it is moving traffic to Azure and splitting core services to improve reliability after nine May incidents caused degraded performance.
Intelligence analysis by GPT-5.4 Mini
The May 2026 availability report pairs GitHub’s infrastructure overhaul with a month of service problems. GitHub says traffic growth, especially from AI-assisted and agentic workflows, is pushing it to add capacity and remove shared failure points.
GitHub is rebuilding parts of its giant machine so one broken piece does not slow everything down, like splitting toys into different boxes instead of one big pile. It still had some May hiccups, but it says the fixes should make the system sturdier as more people use it.
Analysis
GitHub’s monthly report says the company is changing its infrastructure to cope with rapidly growing traffic, which it links in part to AI-assisted and agentic development workflows. The company says it is shifting more load to Azure for elastic capacity, breaking apart the monolith into isolated services, and removing shared failure points that have caused incidents before.
The report gives a few concrete milestones. GitHub says 40% of monolith traffic is now served from Azure, up from 8% in February, with Git traffic at 30% and repository replication at 99%. It also says it has more than doubled effective capacity in four months. On the database side, GitHub says it is separating users, authentication, and authorization into independent domains, and that the new users service is fully cut over and handling twice the traffic at lower database cost. It also says stateless authentication tokens are rolling out to reduce per-request database lookups.
May still included nine incidents with degraded performance. One on May 4 involved a schema migration against a large database table; as normal traffic rose, the combined load saturated connection capacity and caused contention and timeouts across services. GitHub says it paused the migration after detecting impact, and then added plans for tighter scheduling, dynamic throttling, automatic circuit breakers, and earlier alerts tied to migration pressure.
A second set of incidents on May 5 and May 6 affected GitHub Actions hosted runners in East US. The company says the events were connected: remediation after the first incident introduced the configuration issue behind the second. The result included failed or delayed runner requests and about 8,500 Copilot code review requests timing out. The report says GitHub is using these events to refine scaling behavior and reduce the blast radius of future failures.
Key points
- GitHub says traffic is growing quickly, driven in part by AI-assisted and agentic development workflows.
- The company reports major infrastructure work, including a move to Azure and more service isolation.
- In May, GitHub says it saw nine incidents that caused degraded performance.
- A May 4 database migration incident caused latency and request failures across several services.
- Two GitHub Actions incidents on May 5 and May 6 affected hosted runners and Copilot code review requests.
If the migration to Azure and the service split keep working, GitHub should have more room to handle growth without one busy database slowing everything down. The new throttling and circuit-breaker controls could also make routine maintenance safer.
If traffic keeps rising faster than the rebuild lands, routine migrations and runner scaling could still trigger timeouts or failures. The report also shows that shared dependencies can spread a local issue across pull requests, Actions, Copilot, and other services.
