Cloudflare says its “worst outage since 2019” was caused by a failure in its Bot Management system — not a cyberattack or AI-related issue. A change in how its ClickHouse database handled queries created大量 duplicate rows in a configuration file used by the bot-scoring model. The file ballooned past memory limits, crashing the core proxy system that processes customer traffic.
Sites relying on Cloudflare’s bot rules accidentally blocked real users, while those not using bot scoring stayed online. The outage temporarily took down major services, including X, ChatGPT and Downdetector.
Cloudflare says it will prevent similar incidents by hardening config file handling, adding more kill switches, preventing error reports from overwhelming systems, and reviewing failure modes across proxy modules.
Source: The Verge
