In its official report, Cloudflare addressed the widespread outage that began at around 14:50 yesterday. The company explained that the problem was not due to a cyberattack, but rather due to an erroneous change in database access levels that generated a corrupted configuration file for the bot management system.
This corrupted file was distributed throughout the network and, because its size exceeded the allowed limit, the main part of the traffic routing on Cloudflare servers failed and 5xx errors appeared widely.
Cloudflare explained that it initially thought it was facing a major attack, because the system behavior was fluctuating and unusual, and even the company's status page was disrupted at the same time. However, after the source of the problem was identified, the file was stopped being published and the previous healthy version was replaced. Gradually, services returned to normal and by around 20:36 on November 17, all parts were fully restored.
Cloudflare described the disruption as unacceptable and announced that it was strengthening internal mechanisms to prevent similar errors from occurring in the future.