GitHub outage history

Users across GitHub experienced degraded performance and failures with Copilot Code Review beginning around 8:51 PM UTC. GitHub's status updates described a disruption to Copilot Code Review and ongoing investigations into the impact on related services. By 9:01 PM UTC the team had identified the root cause and started applying mitigations, with a rough expectation of recovery within about 30 minutes.

Later, a separate issue emerged around 10:06 PM UTC affecting the Repos Contents API, further impacting users who rely on repository data access. GitHub continued updates and mitigations, and services returned to normal by 10:26 PM UTC. The incident lasted a little over an hour from the initial disruption to full recovery, with Copilot-related disruptions giving way to broader repository data access issues before full restoration.

Users experienced a brief disruption affecting Copilot Code Review, with degraded performance for some GitHub services. The issue began at 8:41 PM UTC and was resolved by 8:46 PM UTC, lasting about five minutes. GitHub described the event as a disruption in Copilot Code Review and said they were investigating reports of impacted performance before restoring normal service.

Users experienced degraded availability for the Grok 4.6 model across Copilot Chat, VS Code, and other Copilot products starting at 2:21 PM UTC. The issue was traced to an upstream Grok model provider, causing Copilot features that rely on the Grok model to perform intermittently or slower than normal.

A follow-up update at 2:26 PM UTC noted that the Grok 4.5 model was affected as well. The problem began to improve over the next few hours, and Grok model availability returned to normal by 5:16 PM UTC, with the Copilot AI Model Providers component returning to normal operation.

GitHub experienced a partial impairment in commit processing that caused pull request diffs to appear stale for several minutes, slowing PR reviews for some users. Impact began at 3:01 PM UTC when the Pull Requests component was flagged as having delays and the team began investigating while scaling up resources. The situation persisted as investigators worked to restore normal update times. 4:01 PM UTC notes indicated the diffs began improving toward normal thresholds, and by 4:06 PM UTC the PR diff updates were back to typical speed, with the component returning to normal.

In short, users experienced slower PR diff updates during this window, impacting review workflows briefly before full operation was restored.

GitHub Copilot users faced degraded performance as Copilot AI Model Providers experienced an elevated rate of errors for OpenAI models, including gpt-5.2, gpt-5.3-codex, gpt-5.4, and the gpt-5.6 family. Other models were not impacted. This led to Copilot-assisted coding suggestions failing or returning unreliable results for some users.

An extended outage affected GitHub Copilot AI Model Providers, degrading availability for the Kimi K3 model used in Copilot products and IDE surfaces. The incident began at Aug 27, 2026 10:06 AM UTC and Copilot features became unreliable as upstream model providers encountered problems, with some pages intermittently inaccessible. The disruption persisted for several hours as engineers worked to restore service.

GitHub's status updates documented ongoing investigation and mitigations, including guidance to switch to another Copilot model or to Auto to continue using Copilot. A concurrent disruption in GitHub Billing was monitored as part of the stabilization efforts. By 7:46 PM UTC, Copilot AI Model Providers had recovered and conditions were stabilizing, with teams continuing to address the root cause to prevent a recurrence.

GitHub faced a broad outage beginning at 11:01 PM UTC, impacting Actions workflow runs triggered by pull request events and leading to elevated delays and timeouts. About 20% of Actions runs were delayed by more than five minutes, and a portion failed to trigger. The incident also affected billing pages and Copilot-related features, with users reporting slowed page loads and failures to initialize Copilot sessions. Pull request merges were blocked as mitigations were applied. Shortly after, Billing disruptions emerged, with users seeing failed billing page loads and related errors as engineers investigated root causes.

Mitigations were applied within the first hours, and recovery began for Actions and PR merges by the early hours of 12:06 AM UTC. While Actions showed signs of recovery, Billing disruptions persisted as teams continued investigation and mitigations. A separate issue affecting Copilot AI Model Providers began at 10:06 AM UTC due to an upstream provider problem, triggering degraded availability for Copilot model services. By 7:46 PM UTC the Copilot components were back to normal and GitHub services resumed a steady state.

GitHub users faced an outage in GitHub Actions starting around 3:16 PM UTC, with Pages showing degraded performance. Workflows failed to trigger or ran into delays, and page hosting experienced slower responses. The impact was most felt by developers relying on automated jobs and continuous deployment.

Users attempting to run GitHub Actions workflows experienced failures and delays as jobs did not start or ran slowly, with some pages showing degraded performance in the early phase. The disruption began around 3:16 PM UTC when the Actions incident was first reported. A database primary failure forced a failover to a replica and throttled inbound traffic, with upstream Vitess issues complicating the recovery. Over the next hour, status updates described a gradual stabilization as engineers worked to regain control of the system.

GitHub Actions experienced startup delays that degraded workflow performance. The issue began at 2:01 PM UTC when GitHub flagged an investigation into delays starting runs. This affected automated workflows, with longer startup times and occasional queuing for Actions. The problem appeared isolated to the Actions component, and service performance returned to normal by 2:26 PM UTC, as the Actions component moved from warn back to up on the status page.