Degraded availability for GitHub Actions, GitHub Pages and Webhooks
Frequent outages in GitHub Actions, Pages, and webhooks are prompting engineers to question GitHub’s reliability and the wisdom of consolidating CI/CD, code hosting, and GitOps on a single vendor. Commenters attribute the instability to a mix of Azure migration challenges, surging AI-driven traffic, and generous free tiers, and debate whether Microsoft is failing to prioritize paid users and architectural fixes. Many are now actively evaluating alternatives such as self-hosted GitLab, Forgejo, or new platforms, while others argue that GitHub’s social network effects and tooling integration make migration costly and complex.
Immediate impact of the outage
- Many reports of GitHub Actions runs stuck in “Queued,” including for self‑hosted runners.
- Some users saw jobs actually complete on their own runners while GitHub’s UI still showed them as queued.
- People wasted time debugging their workflows before realizing it was a platform issue.
- Frustration that the status page updated only after users had noticed and investigated problems.
Vendor lock-in and organizational risk
- Several comments highlight that companies moved “all in” on GitHub Actions and GitOps, creating a single point of failure.
- Some engineers say they warned against this but were ignored; the outages are now seen as management’s problem.
- Others worry that the perceived risk of switching providers, plus GitHub’s network effects, makes migration politically and practically hard.
Alternatives and self‑hosting
- Mentioned hosted alternatives: GitLab, Bitbucket, Cursor Origin, various smaller CI/CD tools.
- Strong preference from some for self‑hosted GitLab, Forgejo, Gitea, etc., often combined with self‑hosted runners.
- Users stress that cloud‑hosted GitLab also has outages; the main reliability gain reported comes from self‑hosting and controlling maintenance windows.
- Some argue GitHub should be treated as a mirror, not the primary source of truth, for critical deployments.
Perceived causes of instability
- One camp blames Microsoft’s acquisition, Azure migration, and corporate culture; they point to historical uptime graphs and insider writeups.
- Another camp emphasizes explosive AI‑driven growth (LLM/agent commits, PRs as “journals”) over the acquisition itself.
- There is debate over whether AI/bot traffic is the main driver versus architectural/organizational problems; evidence is described as partial or unclear.
Free tier, pricing, and multi‑tier service
- Suggestions: sharply reduce free Actions minutes, lower concurrency for free users, or raise prices to fund more capacity.
- Others warn this could drive users to competitors, though some question where credible “sharks” actually are.
- Paying customers resent sharing the same degraded plane as heavy free‑tier and AI usage; some ask for isolation or enterprise‑only instances (e.g., separate hosting options).
Status communication and user coping
- Users want earlier “something is wrong, we’re investigating” messages, ideally driven by automatic detection (e.g., queue depth).
- Some joke about using outages as enforced breaks; others treat recurring outages as a sign it’s time to plan an exit.