Overview
- Cloudflare confirmed the incident was not an attack, with a faulty configuration triggering widespread 5xx errors and login failures across services including ChatGPT, X, Spotify and Uber.
- Engineers traced the failure to a database permissions change that produced an oversized feature file for Bot Management, which propagated across the network and exceeded software size limits.
- The disruption began around 11:20 UTC on Nov. 18, core traffic recovered by roughly 14:30 UTC, and Cloudflare later reported all systems operating normally.
- Multiple products were impaired, including Turnstile, Workers KV, Access authentication and WARP, and Cloudflare’s status and support portals were intermittently unavailable.
- CTO Dane Knecht apologized for the outage, said the configuration mistake was unacceptable, and noted temporary mitigations such as disabling WARP in London during recovery.