An internal constraint on the fleet managing connections to private VPC origins triggered increased 5xx errors for Amazon Web Services CloudFront customers using that connectivity. The disruption began around 12:45 AM PDT on 16 July and persisted until full recovery at approximately 4:18 AM PDT.
The sequence unfolded with precision. When the constraint threshold was reached, the system responsible for distributing routing configuration to network processors failed to load the updated configuration data correctly. This directly impaired routing of VPC Origin connections. Customers relying on other origin types remained unaffected throughout.
Mitigation steps implemented at 3:52 AM PDT restored service less than 30 minutes later. AWS advised those who had temporarily altered their origin type during the incident that they could now revert those changes safely.
The outage halted access to the UK's National Lottery website and mobile app. Hugging Face, the prominent AI platform, became unavailable from most regions. These interruptions, though contained to a few hours, rippled through digital services that millions depend upon for routine operations and critical functions alike.
Such episodes reveal the mechanics of concentrated infrastructure. A single constraint within one hyperscale provider's private networking layer proved sufficient to interrupt connectivity on a global scale. The affected architecture, which permits CloudFront distributions to reach resources inside Amazon Virtual Private Clouds without public internet exposure, had been positioned as a secure default for many deployments.