GitHub Outage History

GitHub’s Pull Requests feature experienced degraded availability, with users reporting difficulties creating new PR threads and adding comments to specific lines and files. Top-level comments on pull requests remained usable, and core PR actions such as opening and merging PRs continued to function. The disruption began at 3:26 PM UTC and generated a global signal as the afternoon progressed, with a notable volume of user reports indicating widespread impact.

An outage briefly disrupted Copilot Cloud Agent sessions on GitHub, preventing users from starting Copilot Cloud Agent sessions or viewing them. The disruption began at 11:26 AM UTC and affected early activity as users attempted Copilot workflows.

GitHub confirmed the disruption and the service was restored by 12:01 PM UTC, after roughly 35 minutes. The issue appeared limited to Copilot Cloud Agent session initiation and viewing, with other GitHub features not indicated as affected. User reports in the late morning aligned with the impact window, before recovery.

The GitHub Actions service experienced degraded availability as Ubuntu standard hosted runners encountered long queues and slow job execution. The incident began at 7:21 AM UTC with early reports of delays and ongoing investigations by the Actions team.

In East US, GitHub Actions experienced degraded performance with elevated queue times and failures impacting a portion of runs. Hosted Runners in the region, including those with private networking, could fail over to a different Azure region to mitigate the issue. Separately, SSH-based Git operations saw elevated latency and failures between 2:00 PM UTC and 4:10 PM UTC UTC, while HTTP-based operations were largely unaffected.

GitHub and its compute partner pushed mitigations and ongoing recovery steps through the afternoon. A mitigation for long queue times was applied around 3:56 PM UTC, with signs of recovery beginning by 4:36 PM UTC as standard Hosted Runners improved and the SSH-related impact was contained. By 6:36 PM UTC the degradation had been mitigated and services were monitored for stability, marking the return toward normal operation.

An outage affected multiple GitHub services, causing degraded performance and partial outages for users. Issues and Webhooks were among the first impacted, with downstream effects on Git Operations, Actions, Packages, Pull Requests, Pages, and Codespaces. In practice, users experienced failed PR reviews (including errors like invalid JSON), automation slowdowns, and slower page loads across the platform.

Status updates tracked the progression: the incident began at 3:46 PM UTC with degraded performance for Issues and Webhooks, followed by Git Operations degradation at 3:51 PM UTC and broader impact to Actions and Packages at 3:57 PM UTC. Pull Requests entered a degraded state at 4:01 PM UTC, and Pages and Codespaces showed issues by 4:06 PM UTC. The platform began recovering in stages, with Git Operations returning to normal by 4:26 PM UTC and full recovery by 4:41 PM UTC.

From 2:01 PM UTC, GitHub faced an outage affecting GitHub Actions with hosted ubuntu-latest and ubuntu-24.02, causing high wait times and job start delays. Other hosted labels and self-hosted runners were not affected initially. The disruption extended to pull request data as Elasticsearch indexing did not reflect all indexed documents, leading to incomplete PR listings and search results on some pages. Users across regions reported timeouts and slow builds during the peak of the incident.

GitHub rolled out mitigations to unblock Actions and began reindexing Elasticsearch indexes. Data remained intact and pages not relying on Elasticsearch stayed usable through the recovery period. Over the following days, indexing progressed and most data and PR listings were restored; by 4:16 AM UTC, the Pull Requests component was up and the incident was substantially resolved.

Impact began at 4:31 PM UTC as GitHub reported degraded search, spreading to core workflows and features. Users experienced workflow run failures, projects failing to load, and timeouts when searching, with pull requests, issues, and packages also affected. The degradation appeared across several components and was noticeable to users around the afternoon and evening, leading to broader service disruption for many developers and teams.

By 7:51 PM UTC, GitHub indicated the root cause was an excessive load on ElasticSearch clusters and disabled the offending source, with signs of recovery soon after. The issue was reframed as having been mitigated by 9:36 PM UTC, though monitoring continued to ensure stability. Recovery culminated with services returning toward normal by 10:51 PM UTC.

Impact began at Apr 24, 2026 7:06 PM UTC when GitHub flagged delays with Actions Jobs for Larger Runners using VNet Injection in the East US region. The issue led to degraded performance for workflows and related services, with updates noting ongoing investigation and a broader public impact tied to Azure East US workloads. Users across regions reported problems ranging from difficulty connecting to Copilot to language model availability issues, indicating the incident affected developer tooling and CI workflows in the affected region.

Users experienced degraded performance with GitHub Pull Requests starting around 7:51 PM UTC as the status page flagged an incident for Pull Requests. A regression in the merge queue behavior affecting squash merges or rebases disrupted PR processing, with some merges possibly occurring incorrectly during the window from 4:05 PM UTC to 8:43 PM UTC UTC.

By 8:56 PM UTC GitHub indicated Pull Requests was operating normally again and that the root cause had been identified with a revert in progress. The service returned to up by 9:46 PM UTC. The incident appears to have affected users attempting merges or squash/rebase operations during the impacted period, with the disruption most evident in merge outcomes during the window above.

GitHub experienced a brief disruption where users could not start Claude and Codex agent tasks from the web. The disruption began at 7:31 PM UTC, as the service entered a degraded state for web-initiated Claude and Codex tasks. Engineers identified the root cause and began mitigation by 7:36 PM UTC. The service returned to normal by 7:46 PM UTC. A single user report was logged during the window at 7:30 PM UTC, indicating limited user impact. The issue appeared limited to the web-based startup flow for Claude and Codex agents, with no broader GitHub services affected.