Gaurav SenOn 18 November 2025, Cloudflare experienced a major global outage.
This is a summary of its incident report.
The failure began when Cloudflare introduced fine-grained user access control its ClickHouse database.
This unintentionally exposed every shard’s feature list as separate rows in their bot check query.
The resultant oversized config propagated across the network, failing the core proxy (which tried to preallocate memory for all features).
Each failure triggered heavy logging, further increasing latency.
Engineers initially assumed this was a DDoS because the failures were intermittent, but later confirmed that the root cause was an internal error.
Incident Report: https://blog.cloudflare.com/18-november-2025-outage/
System Design Course: https://interviewready.io/learn/system-design-course