GitHub's Historic Uptime

Current Outage Context

  • Discussion is prompted by an ongoing outage breaking PR merges, reinforcing the perception of recent instability.
  • Several commenters say they now see issues (e.g., unicorn error pages, flaky clones) often enough to plan around them.

Trends in GitHub Uptime

  • The visualized historical uptime shows a noticeable decline in reliability over recent years.
  • Many feel GitHub was significantly more reliable before the Microsoft acquisition, though others argue usage and complexity have grown so comparisons are unfair.
  • Some note personal self‑hosted setups or small VPSes appear more reliable than GitHub lately.

Role of New Features and Service Scope

  • A major share of downtime spikes is attributed to GitHub Actions, which didn’t exist in earlier “clean” years.
  • Critics argue GitHub has grown from “just a git host” to a large multi‑feature platform, naturally increasing failure surface.
  • Others say even core Git operations now feel less stable, independent of newer features like Actions or Copilot.

Azure Migration and Correlated Outages

  • Several participants link GitHub’s issues to migration to Azure, citing external articles and personal experience with Azure outages.
  • Some report near‑perfect correlation between Azure incidents (e.g., Key Vault issues) and GitHub problems.
  • Azure’s own public status page is viewed as under‑reporting issues.

How Uptime Is Measured and Presented

  • Debate over aggregate uptime metrics: some like a conservative “any subservice down = GitHub down” approach, others find it misleading.
  • Discussion of whether every feature (Pages, Copilot, etc.) should count equally in “GitHub uptime,” and what matters from user vs. enterprise perspectives.
  • Critiques of the chart: truncated y‑axis exaggerating drops, missing feature launch dates, and pre‑2018 periods effectively treated as 100% uptime.

Status Page Accuracy and Historical Data

  • Multiple comments distrust GitHub’s official status page, especially historically, suggesting it was less honest or less instrumented pre‑acquisition.
  • Some suspect improved observability and more transparent reporting, not just worse reliability, explain part of the apparent decline.

Comparisons and Alternatives

  • Bitbucket and Jira are mentioned as having improved over the same period.
  • Others still view GitHub as a valuable, largely free service for open source, even if current reliability is “only” around one or two nines.