Cloudflare outage on July 17, 2020

What this post added

This post details a specific outage event caused by a configuration error in Cloudflare's backbone network. It explains the root cause: an incorrect BGP route leak due to a misconfigured policy-statement that removed a prefix-list condition, leading to all BGP routes being sent to all other backbone routers with a higher local-preference. This caused traffic meant for local compute nodes to be rerouted to Atlanta. The post also outlines the immediate actions taken, including disabling the affected router and restoring service, as well as future preventative measures: implementing a maximum-prefix limit on backbone BGP sessions and changing the BGP local-preference for local server routes.