AWS's Most Popular Region Went Down Because DynamoDB's DNS Had a Race Condition Nobody Had Seen Before
At 3:11 AM ET on October 20, 2025, AWS began receiving alerts that DynamoDB in us-east-1 was failing to resolve. The root cause turned out to be a race condition in DynamoDB's DNS management automation — a latent defect that had existed undetected until a slow Availability Zone caused one of three independent DNS enactors to fall hours behind its peers. The resulting empty DNS record for dynamodb.us-east-1.amazonaws.com didn't just take down DynamoDB — it took down every AWS service that depended on DynamoDB for metadata, control plane operations, or state management. Snapchat, Fortnite, Duolingo, Ring doorbells, and hundreds of banking apps went offline. The technical chain that caused it is one of the most intricate dependency failures in cloud computing history.