Cloudflare withdraws 1,100 customer IP routes via BGP in 6 hours

On 20 February 2026, at 17:48 UTC, Cloudflare experienced a service outage when a subset of BYOIP customers had their routes withdrawn via BGP. The incident was caused by a change in how IP addresses onboarded through the BYOIP pipeline were managed, leading to unintended withdrawal of prefixes rather than a cyberattack or malicious activity.

About 1,100 BYOIP prefixes were withdrawn before engineers could revert the change, and the total duration of the incident was 6 hours and 7 minutes, with most time spent restoring prefix configurations. During the outage, 4,306 of 6,500 prefixes advertised to a peer were BYOIP prefixes, and some customers could restore service by re-advertising their addresses through the Cloudflare dashboard, while ~300 prefixes could not be remediated that way and were manually restored.

The post outlines Code Orange: Fail Small remediation efforts and the steps being taken to better separate operational and configured state, standardise API schema, and improve safeguards to prevent similar outages in the future, according to Cloudflare.