An Update on GitHub Availability

GitHub apologized for two recent outages and unveiled a scaling plan that puts availability first, with capacity and new features following. It began a 10X expansion starting October 2025, but by February 2026 concluded it needed to prepare for 30 times its current scale. Since late December 2025, a sharp rise in agentic development workflows has driven rapid increases in repository creation, PRs, API calls, automation, and work on large repositories.

The pressure from large-scale growth spreads simultaneously across Git repositories, mergeability checks, branch protection, GitHub Actions, search, notifications, permissions, webhooks, APIs, background jobs, caches, and databases. GitHub is analyzing dependencies and traffic tiers to prioritize work, reducing unnecessary work, improving caching, isolating core services, eliminating single points of failure, and moving performance-sensitive paths to systems suited to them.

In the short term, GitHub moved webhooks to a processing layer outside MySQL, and redesigned user session caching and authentication/authorization flows, significantly reducing DB load. The move to Azure also secured additional compute resources.

In the medium-to-long term, GitHub is separating Git and GitHub Actions from other workloads, moving performance- and scale-sensitive code out of its single large Ruby-based codebase into Go, and preparing a path from a single public cloud toward multi-cloud. In response to the spread of large monorepos, GitHub has also strengthened merge queue improvements, and plans to separately unveil a new API design soon aimed at improving efficiency and scale.

On the transparency front, GitHub is adding availability figures to its status page, reflecting both large and small incidents on the status page, and improving how incidents are classified and how customer reports are handled.

• April 23, 2026 merge queue outage: In the merge queue's squash merge, when a merge group contained two or more PRs, an incorrect merge commit was generated. Changes from a previously merged PR and preceding commits were unintentionally reverted in a subsequent merge, affecting 230 repositories and 2,092 PRs. The commits remained intact in Git and there was no data loss, but because the affected default branch state was incorrect, not all repositories could be automatically recovered.
• April 27, 2026 Elasticsearch outage: The Elasticsearch subsystem became overloaded and failed to return search results. The cause may have been a botnet attack, and root cause analysis is still being finalized. There was no data loss, and Git operations and APIs were unaffected, but some PR, issue, and project UI features that depend on search failed to show results.

GitHub stated that these outages are prompting it to focus more on availability, isolation, and reducing blast radius, and that it will continue supporting developers more reliably as an open, scalable platform.