Company

AWS

Every AWS engineering case study on TechLogStack — real production incidents, post-mortems, and fixes.

AWS Distributed Systems
18 min

A Race Condition in DynamoDB's DNS Took Down Snapchat, Fortnite, Ring, and Half the Internet for 15 Hours

It was 11:48 PM PDT on October 19, 2025. Two automation processes inside AWS's DynamoDB DNS management system were doing the same job simultaneously — one fast, one painfully slow. The slow one was just finishing up when the fast one, having already completed, triggered a cleanup job that deleted the slow one's work. In that moment, every DNS record for DynamoDB in the world's busiest cloud region vanished. Snapchat went dark for 375 million daily users. Fortnite lobbies dissolved mid-match. Ring cameras stopped recording. The UK's HMRC tax authority went offline. For 15 hours, the internet's largest database service had no address.

140+ AWS services eventually affected 17M+ outage reports across 3,000+ organizations (Ookla data)