Incident with Github.com [resolved]

Original Article ↗ Hacker News Discussion ↗

Repeated, hours-long outages at GitHub — affecting pull requests, issues, Actions, and the web UI while the official status page lagged or stayed “all green” — are prompting many developers to question its reliability as critical infrastructure. Commenters debate root causes, from Microsoft’s Azure migration and aggressive AI‑driven code generation to underinvestment in core operations, and argue over whether rate limits or higher prices for heavy (often LLM-based) usage are overdue. A growing number of teams report actively migrating or planning to migrate to alternatives such as self‑hosted GitLab, Gitea/Forgejo, or newer federated forges, highlighting broader concerns about centralization, vendor lock‑in, and the fragility of a single dominant code-hosting platform.

Outage symptoms and impact

  • Many users report GitHub being unusable: unicorn error pages, 500s, 404s, “cannot retrieve latest commit,” PR merge status not loading, Actions failing, webhooks not firing.
  • Core git operations often still work (clone, push), and the CLI (gh) or API sometimes works where the web UI does not, but CI/CD pipelines and web-based workflows are blocked.
  • The outage lasts several hours, with some users noting similar incidents just days earlier; some treat it as an enforced break, others as a serious business disruption.

Status page and incident handling

  • The official status page initially shows “all green,” then later “degraded performance” and “~20% error rate.”
  • Many feel this understates the situation: for them it’s effectively 100% down for PRs/issues.
  • Several point out that “degraded performance” incidents don’t seem reflected in uptime stats; older outages are also missing or backfilled as 100% uptime, eroding trust.

Suspected causes

  • GitHub/Microsoft attribute recent availability issues to massive growth, especially AI/LLM-driven code and Actions usage (order-of-magnitude+ increase in commits and CI minutes).
  • Some commenters agree scale is genuinely hard; others argue:
    • Outages predate the AI boom and intensified after the Microsoft acquisition and Azure migration.
    • Other web-scale services manage higher reliability.
    • Over‑feature creep and “vibe-coded”/AI-assisted changes may be harming robustness.

Business, reliability, and SLAs

  • Many organizations now treat GitHub (including Actions) as critical infrastructure; outages block hotfixes, releases, and revenue.
  • Some argue users should have contingency plans, mirrors, and not treat a third-party SaaS as single point of failure; others respond that GitHub explicitly markets itself as reliable enterprise infrastructure with an SLA.
  • There is tension between “don’t rage, go outside” and “we pay for this; frustration is legitimate.”

Alternatives and migration

  • Frequent mentions of moving to:
    • Self-hosted forges (GitLab, Gitea, Forgejo) plus self-hosted CI (Jenkins, Woodpecker, Buildkite, etc.).
    • Other hosted forges (GitLab.com, Codeberg, newer federated options).
  • Tradeoffs discussed:
    • Network effects and social visibility keep people on GitHub.
    • Self-hosting is seen as cheap and reliable by some, operationally burdensome by others.
    • CI/Actions migration and loss of integrations are the hardest parts.

Centralization, community, and pricing

  • Broader reflection that GitHub’s centralization turned distributed git into a single point of failure and a social network for developers.
  • Some welcome a “diaspora” back to more distributed or federated tooling; others fear loss of discoverability and shared norms.
  • Many suggest GitHub should throttle or charge more aggressively for heavy/LLM traffic and free-tier usage, prioritizing paying and human users, but note business incentives may discourage this.