June 2025 Service Degradations at GitHub

Three separate incidents in June 2025 caused degraded performance across GitHub services, affecting Actions, Copilot, and core network routing.

Actions Run Delays and Failures

On June 5, between 17:47 UTC and 19:20 UTC, the Actions service was degraded, resulting in delayed run starts and intermittent job failures. During this 1 hour and 33 minute window, 47.2% of runs experienced delayed starts averaging 14 minutes, and 21.0% of runs failed.

Impact spilled beyond Actions itself: 60% of Copilot Coding Agent sessions were cancelled, and all Pages sites using branch-based builds failed to deploy (Pages serving remained unaffected).

The root cause was a load spike between internal Actions services that exposed a misconfiguration, which in turn caused throttling of requests in the critical path of run starts. The incident was mitigated by correcting the service configuration to prevent throttling. GitHub has updated its deployment process to ensure the correct configuration is preserved going forward.

Copilot Model Outage

From June 12, 17:55 UTC to 21:07 UTC, GitHub Copilot experienced an outage caused by one of its model providers. Gemini models were unavailable, and Claude models saw reduced availability.

Users encountered significantly elevated error rates for chat completions, slow response times, timeouts, and interrupted chat functionality across VS Code, JetBrains IDEs, and GitHub Copilot Chat. The 3 hour and 12 minute incident was mitigated by temporarily disabling the affected provider endpoints to reduce user impact.

GitHub is updating its incident response playbooks for infrastructure provider outages and improving monitoring and alerting systems to reduce detection and mitigation time for similar issues in the future.

Network Routing Issue

On June 17, an internal routing policy deployment to a subset of network devices caused reachability problems for certain network address blocks within GitHub's datacenters. The incident ran from 19:32 UTC to 20:03 UTC, lasting 31 minutes.

Error rates during the incident were as follows:

  • Authenticated github.com UI users: 3-4% error rates
  • Authenticated API callers: 40% error rates
  • Unauthenticated UI and API requests: nearly 100% error rates
  • Actions: 2.5% of runs delayed (average 8 minutes), 3% of runs failing
  • Large File Storage (LFS) requests: 1% errors

The deployment was rolled back at 19:54 UTC, restoring network availability for the affected systems, and normal operations were fully restored by 20:03 UTC. GitHub is expanding its validation process for routing policy changes to prevent similar issues.

Real-time status updates and post-incident recaps are available on the status page. For more engineering details, see the GitHub Engineering Blog.