GitHub Down: The August 17, 2026 Global Outage Explained

On August 17, 2026, GitHub — the Microsoft-owned platform used by around 225 million developers — went down in a widespread outage that disrupted software work worldwide for roughly seven and a half hours. If you hit the pink unicorn "No server is currently available to service your request" page that day, or your pushes, CI runs, and Copilot all failed at once, this is what was happening. The incident is now resolved.

What broke

The outage was broad rather than isolated to one feature. At its peak, GitHub reported roughly a 20% error rate across web experiences and API traffic, and around a 50% error rate on archive downloads and raw repository content. Affected surfaces included the website, Pull Requests, Issues, Actions, Webhooks, API Requests, Git Operations, Pages, and Copilot. Authentication and identity systems were dragged in too — SAML and OIDC authentication, SCIM, and Team Sync were all listed as impacted, which is what made the outage bite even for teams whose primary workflows might otherwise have degraded gracefully.

Timeline (UTC)

• 13:40 — GitHub opens the incident, reporting impacted performance for some services.
• 13:45 — First quantified: roughly a 20% error rate across Pull Requests, Issues, and other experiences.
• 14:04 — Scope widens to ~20% error on web and API traffic, and ~50% on archive and raw repository content.
• 14:24 — SAML/OIDC authentication, SCIM, and Team Sync added to the impact list.
• 14:45–14:58 — Pull Requests, Issues, Actions, and Webhooks escalate from degraded performance to degraded availability; GitHub begins applying mitigations and monitoring.
• ~16:36 — GitHub identifies the problematic component, applies corrective action, and reports strong signs of recovery.
• ~16:59 — Degradation declared mitigated across most services; monitoring continues, with Copilot lagging behind.
• ~21:15 — Incident declared fully resolved, about seven and a half hours after it began, after further work on scattered login failures affecting Copilot and other services.

What caused it

As of now, GitHub has not published a specific technical root cause. During the incident it said it had identified "the problematic component" and applied corrective action — but that component was not named, and no detailed root cause analysis has been released yet. GitHub has stated it will share a full root cause analysis once it has one.

A few things are worth stating clearly, because outages like this attract quick speculation: GitHub did not attribute the incident to any upstream provider, and it did not state any Azure connection. Any claim naming a specific cause — an Azure failure, a database issue, a bad deploy — should be treated as unconfirmed until GitHub's own analysis is published. This article will be updated when that analysis is available.

For broader context, GitHub has spent months under strain from the surge in AI-driven coding. Its CTO noted earlier in 2026 that the company had set out to expand capacity roughly tenfold and then concluded it needed to build for far greater scale still — the kind of pressure that makes large, cascading incidents more likely, even if it isn't the confirmed cause here.

What to do during an outage like this

1. Confirm it's GitHub, not you: check githubstatus.com and @githubstatus. If there's an open incident, the problem is server-side.
2. Don't debug your own setup. Reinstalling tooling, rotating tokens, or reconfiguring git won't help when GitHub itself is failing.
3. Pause automated retries and deploy pipelines so failed jobs don't pile up while error rates are high; resume once status is green.
4. Wait and retry. A server-side outage has no local fix — monitor status and let it recover.

I write up outages and debugging gotchas as I run into them; if that's useful, more are on Netalith or reach me on LinkedIn.