來源GitHub Blog•較早收集於 2h
GitHub 2026 年 5 月可用性報告：效能總結
#platform-stability#devops#service-healthgithubgithub
💡了解平台可靠性趨勢，以更好地管理您的自動化開發與 AI 部署工作流程。
⚡ 30 秒速覽
有什麼變化
2026 年 5 月共發生九起獨立的服務事件
為什麼重要
頻繁的服務降級可能會中斷依賴 GitHub Actions 或儲存庫存取的 CI/CD 管線與自動化 AI 模型訓練工作流程。從業人員應監控這些報告，以評估其開發基礎設施的可靠性。
下一步行動
檢查您 2026 年 5 月的 CI/CD 管線日誌，確認是否有任何失敗的自動化部署與這些 GitHub 服務事件相關。
誰應關注：Developers & AI Engineers
關鍵要點
- •2026 年 5 月共發生九起獨立的服務事件
- •這些事件導致 GitHub 各項服務出現效能下降
- •該報告為平台用戶提供每月的透明度更新
🧠 深度解析
背景與延伸：來自公開資料，非原文內容。引用 15 個來源。
🔑 增強重點摘要
- •GitHub's infrastructure is under significant pressure from rapidly growing traffic, largely driven by AI-assisted and agentic development workflows, which has necessitated a major infrastructure transformation.
- •The company is actively migrating its infrastructure to Azure for elastic capacity, breaking down its monolithic architecture into isolated services, and isolating its primary database cluster to eliminate shared failure points.
- •Specific incidents in May 2026 included GitHub Actions degradations caused by an incorrectly suspended service account, elevated latency on github.com due to a routine online schema migration, and issues with hosted runners stemming from configuration problems introduced during remediation work.
- •GitHub has experienced a notable decline in reliability, with 257 incidents, including 48 major outages, tracked between May 2025 and April 2026, with GitHub Actions being the most frequently affected service.
- •GitHub has acknowledged systemic architectural weaknesses, such as insufficient isolation between components and inadequate backpressure mechanisms, which have allowed localized failures to cascade across the platform.
📊 競品分析▸ Show
| Platform | Availability SLA | Service Credits | Notes |
|---|---|---|---|
| GitHub | >= 99.9% for key services | Not explicitly mentioned for general availability, but post-incident reviews are published. | Focuses on transparency through monthly availability reports. |
| GitLab | 99.9% for Ultimate customers on GitLab.com and GitLab Dedicated | Yes, for Ultimate customers if availability falls below 99.9% (for SaaS purchased/renewed after 2026-01-01). | SLA covers core platform services like Issues, Merge Requests, Git Operations, Container/Package Registry, and API Requests. |
| Bitbucket | 99.9% for Premium plans, 99.95% for Enterprise plans | Yes, financially backed for Premium and Enterprise plan customers if availability drops below the guaranteed threshold. | Covers core features like pull request creation/viewing, Git transactions (HTTPS/SSH), and Pipeline starts (cloud-hosted runners). |
🛠️ 技術深入
- GitHub is undergoing an infrastructure transformation, including a migration to Azure for elastic capacity.
- The platform is being re-architected from a monolith into isolated services to reduce shared failure points.
- Key components like the primary database cluster are being isolated, with user, authentication, and authorization services being split into independent domains.
- A new users service is fully operational, handling double the traffic at a lower database cost.
- Stateless authentication tokens are being rolled out to reduce per-request database lookups, which can amplify pressure during traffic spikes.
- GitHub utilizes MySQL, scaled with Vitess for horizontal and vertical sharding, to manage its database operations.
- The company is implementing improvements such as decoupling critical services, enhancing load-shedding capabilities, and improving traffic management to bolster reliability.
- GitHub's infrastructure is globally distributed, employing Points of Presence (POPs) for latency reduction, isolated data centers with redundancy, and mechanisms for failover and disaster recovery.
- Incident response follows a structured process including preparation, detection and analysis, containment and eradication, recovery and restoration, and post-incident analysis.
🔮 前景展望基於引用來源的 AI 分析
GitHub will likely continue to experience intermittent service degradations as its infrastructure transformation progresses.
The ongoing migration to Azure and the decomposition of its monolith are complex, large-scale changes that inherently carry risks of new issues arising during the transition, as evidenced by incidents caused by remediation work.
The increasing reliance on AI-driven development workflows will necessitate further significant investments in GitHub's underlying infrastructure capacity and resilience.
GitHub explicitly states that rapidly growing traffic, largely from AI-assisted workflows, is a primary driver of current reliability challenges and is pushing them to transform their infrastructure.
GitHub's incident reporting and transparency will become even more critical for user trust and platform adoption.
With a history of increasing incidents and public developer frustration, transparent communication about outages and ongoing reliability efforts is essential to maintain confidence in the platform.
⏳ 時間線
2020-07
GitHub introduced its monthly Availability Report for transparency on service performance.
2023
GitHub experienced 94 total incidents, including 22 major and 72 minor outages.
2024
GitHub experienced 119 service incidents, including 26 major and 93 minor disruptions.
2025-05
Beginning of a period (May 2025 - April 2026) where GitHub tracked 257 incidents, with 48 major outages.
2026-02-09
GitHub experienced a major outage affecting most services, including GitHub.com, API, Git over HTTPS, Actions, and Copilot.
2026-04
GitHub overhauled its status page, adding a 'Degraded Performance' tier, and Mitchell Hashimoto moved his Ghostty project off GitHub due to reliability concerns.
📎 來源 (15)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
AI 週報
閱讀本週精選 AI 大事摘要 →
👉相關動態
AI 策展新聞聚合。所有內容版權歸原始發布者所有。
原始來源： GitHub Blog ↗
每週電子報
每週一封，可隨時退訂。