The Typo That Took Amazon S3 Down for 4 Hours and 17 Minutes [c68MMMGGmxh]
On February 28th, 2017, an Amazon engineer fixing a slow billing system typed one routine command with one wrong input, and accidentally pulled the index, the filing clerk that knows where every object lives, out of the heart of S3. Half the apps on the internet broke at once. And Amazon couldn't even turn its own status page red, because the red icons were stored in S3. This is the full post-mortem: the typo, the filing clerk that hadn't taken a day off in years, the four-hour restart, and why status pages don't live on their own infrastructure anymore. Chapters: 0:00 One small typo 0:22 One system, millions of logos 0:38 The index: S3's filing clerk 1:03 The command reaches the heart 1:26 The internet finds out what runs on S3 1:48 The dashboard that stayed green 2:09 The slowest restart in years 2:31 The damage 2:46 What the post-mortem found 3:09 Next episode Sources: AWS's official post-event summary (aws.amazon.com/message/41926), contemporaneous reporting; $150M figure is Cyence's estimate for S&P 500 companies. Music: "The Descent" by Kevin MacLeod (incompetech.com), licensed under Creative Commons: By Attribution 4.0, creativecommons.org/licenses/by/4.0/ Narration is AI-generated (ElevenLabs). #aws #s3 #cloudcomputing #postmortem #devops