Cloudflare, which caused a massive access outage for internet services worldwide, said the error on the 18th stemmed from an internal issue triggered by a change in permissions for the databases system, and noted it was unrelated to a cyberattack. The company said core traffic has already returned to normal and pledged to put in place measures to prevent a recurrence.
According to the industry on the 19th, Matthew Prince, Cloudflare CEO, said on the company blog that the network outage began at 11:20 a.m. Coordinated Universal Time (8:20 p.m. Korea time), that most core traffic was restored by 2:30 p.m. (11:30 p.m. Korea time), and that all services returned to normal at 5:06 p.m. (2:06 a.m. the following day, Korea time).
For about six hours, major global services including ChatGPT, Google, YouTube, and X experienced successive access issues due to the outage. Because roughly one-fifth of global internet traffic passes through the Cloudflare network, the system error immediately led to widespread disruptions.
Prince said, "As one of the permissions in the databases system was changed, multiple items were output to a function file used by the management system, causing the problem." He added, "After identifying the issue, we immediately replaced it with the previous version of the file to normalize traffic processing functionality."
Cloudflare apologized for causing inconvenience across the internet ecosystem and said it would prepare recurrence prevention measures, including expanding the global kill switch and applying a file collection framework identical to the user input method. The company plans to soon release an error report and propose ways to prevent system overload.
Prince said, "It was the worst service outage since 2019," adding, "We will build a system with greater resilience to ensure a stable flow of traffic even during failures."