GitHub Down Again?
Frequent 503/504 errors and slow release downloads on GitHub, despite its status page initially showing all systems green, left many users with broken CI pipelines and unable to access critical tools and source code. Commenters argue that the industry’s heavy reliance on GitHub as a single point of failure is increasingly risky, especially as AI-driven activity appears to be straining its infrastructure. Alternatives like self-hosted Git, GitLab, Forgejo, and Codeberg are weighed, but concerns over reliability, features, terms of service, and data use for AI training make migration a complex trade-off rather than an obvious solution.
Outage symptoms and impact
- Many report 503/504 errors, especially for repository pages, commits, and release downloads.
- Some can access GitHub only when logged in or only from certain browsers, suggesting CDN or auth-related issues.
- CI pipelines and Terraform runs fail because they cannot download releases, tools, or GitHub API responses.
- Downstream effects include failure to build Linux/Nix/Elixir projects and to download Linux distro ISOs and utilities that are hosted on GitHub.
Status page and incident reporting
- Several note that githubstatus.com showed “all green” while services were clearly failing.
- Status updates are perceived as delayed, downplaying issues as “degraded performance for some users” even when the web UI feels completely down.
- A third‑party status dashboard is mentioned but is also limited because it relies on GitHub’s own incident data and is only updated daily.
CI choices and self‑hosting
- Some argue companies should host their own CI and avoid large GitHub Actions bills.
- Others counter that opportunity cost, operational overhead, and blame/risk make managed CI attractive.
- A hybrid pattern is described: self-hosted infra for primary builds, with GitHub used as a public mirror for provenance and attestation.
Alternatives to GitHub
- Alternatives discussed: GitLab, Buildkite, Forgejo, Codeberg, self‑hosted bare Git, and more experimental approaches (e.g., Nostr-based git).
- Codeberg/Forgejo are praised as “free software” options but criticized for weaker uptime, slower CI, anti-scraping measures, and missing features.
- Some prefer GitHub for scale and reliability despite outages; others see repeated incidents as a reason to migrate.
AI, traffic, and policies
- Outages are linked by some to surging commit and traffic volume from AI agents.
- Git services are seen as increasingly vulnerable to automated load.
- Codeberg’s policy against repositories “mostly” generated by LLMs sparks debate:
- Supporters say it protects copyright clarity and prevents misuse and scraping.
- Critics argue it restricts user freedom, is hard to enforce reliably, and is unsuitable for serious business use.
Centralization, lock‑in, and resilience
- Heavy ecosystem dependence on GitHub (source, releases, CI, tooling) is viewed as a single point of failure.
- Stars, network effects, and perceived reliability are seen as the main lock‑in forces.
- Some insist engineers should design around such single points (mirrors, local caches), while others accept GitHub’s dominance pragmatically.