GitHub Availability Report: August 2026 Overview

GitHub August 2026 Availability Report Summary

GitHub published its availability report for August 2026, outlining platform performance, infrastructure updates, and several notable service incidents. While the platform continues to experience significant growth and progress in its migration to Azure, August proved to be a challenging month for availability. GitHub reported ongoing investments in architectural improvements, capacity monitoring, retry policies, and core service resiliency.

Infrastructure and Migration Milestones

GitHub shared several updates regarding its ongoing database and cloud migration efforts:

• Azure Database Migrations: On August 11, GitHub ran a production MySQL primary from Azure for the first time with minimal client-observed write impact. Two additional primaries were transitioned on August 27, with more scheduled in the coming weeks.
• Traffic Records: Reads from migrated services peaked at 60.4%, while reads from GitHub’s monolith in Azure reached 64.3%. Git reads reached 54%.
• Query Optimizations: Moving the 24-table authentication-core cohort off the shared database `mysql1` removed approximately one million queries per second from its replicas, while query-hygiene changes eliminated another 120,000 queries per second.
• GitHub Actions Capacity: Job-routing changes moved 33% of jobs from a constrained production cluster to spare capacity, reducing peak cache CPU utilization from 98% to 80% and adding an estimated three months of headroom.

Key August 2026 Incidents

GitHub detailed multiple incidents that affected platform availability during the month:

• August 6 (Actions Service Incident): Lasting 10 hours and 42 minutes, a routine deployment to an internal GitHub Actions service briefly reduced capacity in one data center. The remaining sites exceeded their limits as traffic shifted, causing CPU throttling, memory restarts, and cascading cache, DNS, and API errors. A latent bug further delayed recovery by trapping runners in a retry loop for revoked jobs.
• August 7 (Load Balancer and Authentication Latency): Lasting 7 hours and 35 minutes, a traffic peak pushed one data center's load balancers past their limits. A service-mesh sidecar failed to scale up, exhausting network flow limits and degrading the shared authentication path, which impacted Copilot Token Service and other routing services.
• August 20 (Copilot Cloud Agent Lag): Lasting 9 hours and 54 minutes, a provider-side regional outage affected the managed cloud database storing Copilot cloud agent task status and results. Although tasks ran to completion and no work was lost, streaming processors fell behind due to database latency, and slow regional failover prolonged the delay.
• August 26 (Actions Shared Database Saturation): Lasting 2 hours and 50 minutes, a burst of incoming events combined with high load pushed a shared database past its tipping point. The primary database saturated, preventing the internal service from assigning runners and causing actions runs to queue or fail.
• August 27 (Kimi K3 Model Disruption): Lasting 2 hours and 8 minutes, an upstream model provider experienced a serving degradation that caused a large share of Kimi K3 requests to fail. Customers using other models or the Auto setting were unaffected.

Practical Implications and Verification

Users experiencing lingering queue delays or workflow disruptions can verify service status through GitHub's official status page. Developers should review retry behaviors in custom applications and workflows to prevent unintended request amplification during regional degradations. Further technical details and engineering insights can be explored directly on the GitHub Blog.