Two December Incidents Degraded GitHub Services
GitHub reported two separate incidents in December 2024 that caused degraded performance across its platform. Neither resulted in data loss, but both affected user-facing experiences to varying degrees.
Live Updates Failure Overloads Web Servers
On December 17, between 14:33 and 14:50 UTC, users encountered intermittent errors and timeouts while accessing github.com. The error rate averaged 8.5% and peaked at 44.3% of requests. The outage had a broad impact, preventing users from logging in, viewing repositories, opening pull requests, or commenting on issues.
The root cause traced back to planned maintenance that accidentally disabled the live updates service. This service powers automatic page updates in the user experience. With it down, clients reverted to aggressive manual refreshing, which overwhelmed the web servers already under strain from the maintenance changes.
Engineers mitigated the incident by rolling back the maintenance changes and scaling up capacity to absorb the influx of WebSocket client traffic. Post-incident analysis revealed gaps in alerting caused by the overloaded servers, which masked the true scope of the impact until after the issue was resolved.
To prevent a recurrence, GitHub is adding monitoring higher in the request path to catch this class of failure earlier and is improving alerting to better gauge the breadth of impact during incidents.
Third-Party Outage Takes Down Marketing Pages
A second incident occurred on December 20, from 15:57 to 16:39 UTC, when some marketing pages became unavailable due to a partial outage at a third-party service provider. Visitors to the affected pages received 500 errors. No operational product or service areas were impacted.
The provider resolved the outage at 16:39 UTC, restoring access. GitHub is now investigating ways to improve error handling and graceful degradation so these pages remain reachable during future third-party disruptions.
