GitHub’s reliability gains in 2026: How infrastructure upgrades cut downtime

GitHub’s reliability efforts are gaining momentum as the platform transitions from a monolithic architecture to a distributed system built for resilience. With AI-assisted development reshaping workflows, GitHub has accelerated its migration to Azure, isolating core services to prevent widespread outages. The results speak for themselves: a 40% reduction in monolith traffic handled by Azure, a doubling of effective capacity in four months, and a 99% uptime rate for repository replication.

Breaking down GitHub’s infrastructure transformation

GitHub’s journey toward reliability isn’t just about scaling—it’s about rethinking how the platform operates at its core. The shift from a single monolithic system to isolated microservices is designed to isolate failures and prevent cascading outages. Three strategic moves are driving this change:

• Azure migration: Over 40% of GitHub’s monolith traffic now runs on Azure, a significant jump from just 8% in February. Git operations account for 30% of this migration, while repository replication reaches 99% availability. This migration isn’t just about capacity—it’s about decoupling services to reduce shared failure points.

• Database isolation: GitHub is splitting its primary database cluster into independent domains for users, authentication, and authorization. This separation ensures that a failure in one area doesn’t spill over into others, a lesson learned from past incidents where database contention caused widespread disruptions.

• Stateless authentication: The platform is rolling out stateless authentication tokens to eliminate per-request database lookups, which previously amplified pressure during traffic surges. This change not only improves performance but also reduces database costs by handling double the traffic with fewer resources.

The guiding principle behind these efforts is clear: availability first, capacity second, features third. GitHub acknowledges that challenges remain, but the focus on structural reliability is reshaping how the platform operates under pressure.

May’s incidents: Lessons from nine disruptions

Despite progress, GitHub faced nine incidents in May that led to degraded performance across multiple services. While no outage was catastrophic, each incident provided valuable insights into where further improvements are needed.

The May 4 database migration mishap

On May 4, 2026, GitHub experienced a service disruption lasting 55 minutes, triggered by a routine online schema migration. The incident affected a broad range of services, including pull requests, issues, actions, and Git operations. At its peak, 1.3% of requests returned 5xx errors, with an average of 0.46% across the duration.

The root cause? A combination of a large, heavily accessed database table and a migration running during peak traffic. As traffic ramped up, the combined load saturated database connection capacity, leading to query contention and cascading timeouts across dependent services. The incident was mitigated within 33 minutes after the migration was paused, with full resolution following shortly after.

Key takeaways:

• Migrations against high-traffic tables now require tighter alignment with low-traffic windows.
• Dynamic throttling will adapt migration pace to live cluster load.
• Automated circuit breakers will pause migrations if latency or connection utilization crosses safe thresholds.

GitHub Actions: Two related outages in East US

May 5 and 6 saw two related incidents affecting GitHub Actions hosted runners in the East US region. The first incident, lasting 3 hours and 49 minutes, saw 13.5% of standard runner jobs and 16% of larger runners fail or delay. Copilot code review requests were also impacted, with 8,500 requests timing out.

The second incident, lasting 2 hours and 25 minutes the next day, was a direct result of remediation work from the first outage. A scale-up operation for hosted runner VMs hit an internal rate limit, triggering a configuration issue that cascaded into further failures.

What’s being done:

• Reviewing rate limits to prevent overload during VM creation.
• Strengthening backoff logic to ensure it triggers correctly under all response codes.
• Improving automated monitoring to catch issues before they escalate.

The road ahead: Reliability as a continuous commitment

GitHub’s infrastructure upgrades are far from complete, but the progress is undeniable. The platform is moving toward a future where failures are isolated, migrations are safer, and user experience remains consistent even under heavy load. With AI-assisted development driving unprecedented traffic, GitHub’s focus on resilience ensures it can scale without sacrificing reliability.

For developers, this means fewer disruptions, faster performance, and a platform that can keep up with the demands of modern workflows. The work ahead will involve refining these systems, but the foundation is already in place. GitHub’s commitment to availability-first design sets a new standard for platform stability in an era of rapid technological change.