GitHub Publishes a Postmortem for Its August 17 Outage
A capacity shortfall in one data center cascaded into a nearly eight-hour outage touching almost every core GitHub service.
GitHub published a postmortem on August 20, 2026 for an outage three days earlier that took down github.com, authentication, GitHub Actions, its APIs, pull requests, issues, and Copilot for roughly 7 hours and 47 minutes worldwide.
According to the postmortem, the root cause was a capacity failure: a critical infrastructure component in GitHub's Central US data center failed to scale as traffic hit a new peak. GitHub noted that its platform has grown from 1.4 billion to 2.9 billion monthly commits since April 2026, and that growth had been putting sustained pressure on the systems involved. The initial capacity failure cascaded into authentication failures across the platform, and during recovery, client-side retry loops in Copilot's services compounded the load rather than backing off, extending the outage.
GitHub was direct in the postmortem about the outage's severity, acknowledging that "a prolonged outage like this one has a real impact on your productivity and on your trust in us," and noting this was the second significant outage the platform had experienced in August.
The remediation plan GitHub laid out covers several fronts: adding capacity and continuing an in-progress migration to Azure infrastructure, which GitHub said already handles 58 percent of platform load; reducing architectural bottlenecks; and putting consistent retry limits and budgets in place across services so that one service's retries during an incident don't amplify load on another already-struggling service, the specific failure mode that extended this outage's recovery time. GitHub also said it would review alerting coverage for infrastructure components that are vulnerable during sudden traffic spikes, and invest further in rollout safety and observability practices generally.
GitHub said further, more detailed root-cause analysis would be included in its regular monthly availability report. The company has published these standalone incident postmortems and monthly availability summaries as its standard practice for communicating outages, giving customers a documented account of what broke and what changes are meant to prevent a repeat.
Source: The GitHub Blog — The August 17 outage, and the work ahead