GitHub Actions and Pages are experiencing degraded availability

Original Article ↗ Hacker News Discussion ↗

Frequent, hours-long outages of GitHub Actions and Pages are leading many developers and companies to question GitHub’s reliability, especially as these services sit in the critical path for CI/CD and deployments. Commenters speculate about causes ranging from Azure migration and explosive AI-driven load to organizational and cultural issues at Microsoft, while noting that official uptime figures and SLAs feel disconnected from their experience. As a result, a noticeable number are exploring or migrating to alternatives such as self-hosted GitLab, Forgejo, Woodpecker, or custom CI setups to reduce dependence on GitHub’s control plane.

Reliability and Uptime Concerns

  • Many commenters say GitHub Actions and Pages outages have become frequent, with users joking there’s effectively “one nine” or “zero nines” of uptime.
  • Third‑party uptime trackers cited show ~94–98% availability over recent periods, versus GitHub’s own much higher SLA numbers, leading to accusations that official metrics are misleading.
  • The length of this incident (around a full workday for some) and multiple incidents in recent months drive a sense that reliability is getting worse, not better.

User Impact and Frustration

  • Work is blocked: CI not running, PRs can’t merge, hotfixes and customer releases delayed, production incidents harder to handle.
  • Error messages are described as misleading or opaque, e.g., failures that don’t clearly indicate “no runners available.”
  • Some enterprises feel they get no extra stability despite paying, and that support/SLA credit mechanisms are weak or “scummy” by design.

Suspected Causes

  • Two dominant theories:
    • Massive load growth, especially from AI agents and Copilot usage, leading to 10–14× more commits and >4× more Actions minutes, stressing architecture.
    • The ongoing migration from GitHub’s old infrastructure/AWS to Azure, with Azure described as flaky and politically mandated.
  • Many believe both load and migration, plus aggressive feature rollout (especially AI/Copilot), are interacting badly.
  • Some argue these problems reflect poor leadership, rushed timelines, and “performative” engineering culture.

Actions Architecture and Self‑Hosted Runners

  • Users are surprised that self‑hosted runners fail too, because the central scheduler and webhooks are down.
  • Several criticize the design: centralized control plane as single point of failure, YAML‑heavy workflows, and poor degradation strategies (no clear load shedding or priority for paying customers).

Alternatives and Self‑Hosting

  • Strong interest in moving CI off GitHub: self‑hosted Jenkins, GitLab, Forgejo, Gitea, Woodpecker, Buildkite, Argo, and various niche CI systems are mentioned.
  • Some report good experiences with self‑hosted GitLab/Forgejo + runners (often on a single server), claiming better reliability and control.
  • Others note vendor lock‑in and switching costs mean GitHub still “wins” despite outages, especially due to its PR/review UX and network effects.

Broader Commentary

  • Threads connect GitHub’s issues to wider concerns: AI “slop,” RAM/datacenter pressure, Microsoft’s product quality, and the risks of centralizing critical dev infrastructure in one proprietary platform.