GitHub pledges architectural overhaul to scale after second August outage

GitHub has announced a multi-part plan to improve platform reliability after two outages in August disrupted Actions, pull requests, issues, Copilot and APIs. The core of the plan is a new architecture that scales read capacity linearly with the number of readers, beginning with the largest monorepos. In parallel, GitHub is accelerating the shift of workloads to Microsoft Azure, which already handles a large share of the platform’s Git traffic. Additional measures include tighter retry limits, isolation of critical systems, and alerts for early traffic spikes.

Adopting these changes will require engineering teams to review how their workflows interact with GitHub’s APIs and to verify compatibility with the revised retry limits. Moving workloads to Azure may involve transferring data, adjusting build agents, and validating performance under the new load distribution. Teams should observe the system during the rollout to detect any unexpected behavior as traffic shifts. Organizations that depend heavily on GitHub Actions for continuous integration will need to allocate time for testing and adjustment.

The new read-scaling architecture will only improve reliability if it truly scales linearly with reader count; any shortfall will leave the same bottlenecks in place. Moving more work to Azure assumes the cloud can absorb the additional load without introducing new failure points. Tightening retry limits must be balanced so that legitimate requests are not inadvertently blocked. Isolating critical systems helps contain failures, but any oversight in the isolation boundaries could let problems spread. Finally, alerts for traffic spikes provide early warning only when they are set to fire before overload occurs.