{"UUID":"402733d3-31ee-4ca3-b769-0a9cdf7dd59f","URL":"https://github.blog/news-insights/company-news/github-availability-report-february-2026/","ArchiveURL":"","Title":"GitHub Actions and Codespaces outage of February 2026","StartTime":"2026-02-02T18:35:00Z","EndTime":"2026-02-03T00:30:00Z","Categories":["automation","cascading-failure","config-change","security"],"Keywords":["actions","codespaces","copilot","dependabot","codeql","github pages","vm","telemetry"],"Company":"GitHub","Product":"GitHub Actions, GitHub Codespaces, GitHub Copilot, CodeQL, Dependabot, GitHub Enterprise Importer, GitHub Pages","SourcePublishedAt":"2026-03-12T03:23:54Z","SourceFetchedAt":"2026-05-04T19:51:26.790455Z","Summary":"A telemetry gap caused security policies to be auto-applied to backend storage accounts in GitHub's underlying compute provider, blocking access to critical VM metadata. All VM create/delete/reimage operations failed, taking out Actions hosted runners in every region, Codespaces, Copilot coding agent, CodeQL, Dependabot, GitHub Enterprise Importer, and Pages for ~5h53m.","Description":"On February 2, 2026, GitHub experienced a significant outage affecting multiple services for approximately 5 hours and 53 minutes. The incident began at 18:35 UTC, with services gradually recovering, and full recovery for larger runners achieved by 00:30 UTC on February 3, 2026.\n\nThe outage primarily impacted GitHub Actions hosted runners and GitHub Codespaces, rendering them unavailable. Other critical GitHub features relying on the same compute infrastructure were also affected, including Copilot coding agent, Copilot code review, CodeQL, Dependabot, GitHub Enterprise Importer, and GitHub Pages. The impact was widespread, affecting all regions and runner types, with Codespaces creation and resume operations failing globally.\n\nThe root cause was identified as a loss in telemetry, which led to the mistaken application of security policies to backend storage accounts within GitHub's underlying compute provider. These policies inadvertently blocked access to critical VM metadata, causing all virtual machine operations such as creation, deletion, and re-imaging to fail.\n\nMitigation involved rolling back the erroneous policy changes, which commenced at 22:15 UTC. As virtual machines came back online, the system processed the backlog of requests. GitHub is actively collaborating with its compute provider to enhance incident response, improve early detection mechanisms, and ensure safer rollouts of similar changes in the future."}