On November 18, 2025, Cloudflare experienced a significant outage that affected large portions of the internet, including services like Shopify, Amazon, ChatGPT, Spotify, and Reddit. This three-hour disruption was rooted in a database permissions change within Cloudflare’s ClickHouse cluster, which inadvertently caused the bot management system’s feature file to double in size due to duplicate entries. This increase exceeded the configured size limits in the proxy system, resulting in widespread 5xx errors.
The problem was compounded by a faulty query that repeatedly generated corrupted configuration files every five minutes, leading to intermittent failures across Cloudflare’s network. Initially, the unpredictable nature of these failures led engineers to suspect a Distributed Denial of Service (DDoS) attack. Indeed, the sudden drop in service visibility, as even Cloudflare’s status page was affected, added to this confusion.
Ultimately, engineers were able to identify the root cause: the database query’s failure to respect the intended data structure following the permissions change. Recovery involved halting the generation of the faulty files, manually deploying a known-good configuration, and restarting affected services. Full resolution was achieved by 17:06 UTC after a series of emergency procedures, including stabilizing in the failing state, fixing file ingestion issues, and implementing more robust validation checks.
This incident underscores the risks associated with internet centralization around a few key infrastructure providers like Cloudflare. The outage has prompted discussions on the necessity of diversified architectures and multi-cloud strategies to mitigate the impact of similar incidents in the future. Cloudflare has committed to enforcing stricter validation protocols, developing global kill switches, and extending comprehensive failure mode reviews to prevent recurrences of such disruptions.
The high-profile outage serves as a stark reminder of the vulnerabilities inherent in today’s internet infrastructure and highlights the critical need for transparency and robust post-incident analysis.

Collections

Cloudflare experienced a three-hour outage on November 18, 2025, affecting major services like Shopify, Amazon, and ChatGPT. A database permissions change in their ClickHouse cluster caused the bot management feature file to double in size with duplicate entries, exceeding proxy system limits and triggering widespread 5xx errors. A faulty query generated corrupted configuration files every five minutes, creating intermittent failures that initially appeared as a DDoS attack. Engineers resolved the issue by halting faulty file generation, deploying a known-good configuration, and restarting services. Cloudflare committed to stricter validation protocols, global kill switches, and comprehensive failure mode reviews to prevent future incidents.

Cloudflare Outage on November 18, 2025: A Database Permissions Mishap