GitHub is once again down
Frequent outages at GitHub, including a new incident affecting pushes and pull requests, are fueling frustration among developers who rely on it for core workflows and CI/CD. Many attribute the instability to GitHub’s migration to Microsoft Azure and its heavy focus on AI features like Copilot, arguing this has come at the expense of reliability and maintainability. As trust erodes, users are increasingly contemplating or adopting alternatives such as self-hosted Forgejo, Gitea, Gerrit, or GitLab mirrors to reduce dependence on a single, fragile platform.
Overall reaction & user impact
- Many commenters express strong frustration; several say outages now feel weekly or biweekly.
- People report 500 errors when pushing branches, creating PRs, or running GitHub Actions; some have critical deploys blocked.
- Some say status page often claims “all good” while jobs remain stuck.
- A few accept occasional downtime as tolerable for git hosting, but not for CI/CD and production pipelines.
Perceived causes of instability
- A major theme: the ongoing migration from GitHub’s own data centers to Azure.
- Cited memo: GitHub is capacity‑constrained in Virginia and prioritizing a full Azure move within ~24 months, overlapping old/new infra for at least six months.
- Several argue any “massive infra migration” doesn’t have to cause this much pain, calling this a botched migration.
- Others blame increased AI load: Copilot, “vibe coding,” and AI agents causing far more automated traffic.
Historical reliability & Microsoft’s role
- Mixed memories: some recall GitHub being flaky even pre‑acquisition; others say outages have clearly worsened since ~2023.
- There’s debate on whether this is mainly “Microsoft culture” (Azure-first, AI-first, cost-cutting, leadership changes) or a continuation of long‑standing issues.
- Some note key leadership and senior engineers have left, which is seen as contributing to decline.
Alternatives, mitigations, and decentralization
- Multiple users mention moving to or evaluating Forgejo, Gitea, GitLab, Gerrit, or self‑hosted solutions; mirrors on other platforms are common.
- Several advocate remembering “GitHub is not git” and promoting distributed workflows and backups.
- Git mirrors and local clones are cited as partial mitigation when GitHub is unavailable.
Architecture, reliability practices, and “nines”
- Light discussion of uptime “nines” and jokes about how far GitHub is from high availability.
- Some criticize microservices sprawl and complex distributed systems, contrasting them with simpler monoliths.
- There’s debate over why GitHub doesn’t do safer rollouts, blue/green deployments, or full duplicate stacks with realistic traffic replay; others counter that such retrofits are enormously difficult in a legacy system.
AI focus and product direction
- Many see GitHub’s AI push (Copilot, “vibe coding”) as distracting from its core role: reliable git hosting and collaboration.
- Some argue that chasing AI features and shareholder growth has crowded out investment in stability and user experience.