Best of Cloudflare — November 2025

  1. 1
    Article
    Avatar of cloudflareCloudflare·27w

    Replicate is joining Cloudflare

    Cloudflare acquires Replicate, a platform for running AI models with 50,000+ models in its catalog. The integration will bring Replicate's model catalog and fine-tuning capabilities to Cloudflare's Workers AI platform, while maintaining existing APIs for current users. The combined platform aims to provide serverless GPU inference on Cloudflare's global network, unified model management through AI Gateway, and seamless integration with Cloudflare's developer tools including Workers, R2, Vectorize, and Durable Objects.

  2. 2
    Video
    Avatar of fireshipFireship·27w

    The entire internet just crashed... again

    Cloudflare experienced a major global outage affecting millions of websites including ChatGPT, X (Twitter), and League of Legends. The root cause was a latent bug in their bot mitigation service triggered by a routine configuration change. An automatically generated configuration file for managing threat traffic grew beyond its expected size, causing crashes across multiple Cloudflare services. The incident highlights the internet's dependence on centralized infrastructure providers and the cascading impact when they fail.

  3. 3
    Video
    Avatar of awesome-codingAwesome·27w

    The whole internet was down... again...

    Recent major outages from Cloudflare and AWS exposed critical vulnerabilities in modern internet infrastructure. While cloud services promised decentralization and resilience, the industry has consolidated around a few vendors using default configurations. Cloudflare's outage was caused by an oversized feature file in their Bot Manager component. The real issue isn't the outages themselves, but the illusion of resilience created by cloud-native tools while actually centralizing failure points. Modern developers increasingly lack the knowledge to build systems that gracefully handle failures, relying instead on configuration wizards and AI assistance.

  4. 4
    Article
    Avatar of collectionsCollections·29w

    Open-Source Release: tokio-quiche Simplifies Async QUIC and HTTP/3

    Cloudflare open-sourced tokio-quiche, an asynchronous QUIC library for Rust that combines quiche protocol with Tokio runtime. The library simplifies building QUIC applications by providing async UDP socket integration, connection management, and HTTP/3 support out of the box. It uses an actor model architecture and supports custom protocols through traits. Already powering Cloudflare's production infrastructure including iCloud Private Relay and Oxy proxies, handling millions of HTTP/3 requests per second, the library is now available for developers to build QUIC-based applications in the Rust ecosystem.

  5. 5
    Video
    Avatar of lowlevellearningLow Level Learning·26w

    did rust take down the internet?

    Cloudflare experienced a major outage caused by a Rust code issue involving the unwrap() function. When their bot management system received a configuration file exceeding the hardcoded 200-feature limit, the unwrap() call panicked and crashed services. The incident sparked debate about Rust's error handling semantics, particularly whether unwrap() should be renamed to better convey its panic behavior. The root cause was improper error handling in production code rather than a fundamental language flaw.

  6. 6
    Video
    Avatar of primeagenThePrimeTime·27w

    So cloudflare went down

    Cloudflare experienced a major outage caused by their bot management system receiving over 200 feature flags instead of the expected 60. The service used Rust's unwrap method for error handling, which panicked when the pre-allocated memory limits were exceeded. The incident highlights the trade-offs between performance optimization through pre-allocation and robust error handling in high-traffic server environments. While some criticized Rust, the failure would likely have occurred regardless of language choice given the architectural constraints.

  7. 7
    Article
    Avatar of collectionsCollections·27w

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

    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.

  8. 8
    Article
    Avatar of theregisterThe Register·26w

    70-hour work weeks no longer enough for Infosys founder

    Infosys co-founder Narayana Murthy advocates for 72-hour work weeks, praising China's controversial 996 culture despite legal challenges and worker protests. In other Asia-Pacific tech news: Tokyo court orders Cloudflare to pay $3.3M for enabling manga piracy through CDN services; India and Europe plan to link payment systems (UPI and TIPS) for cross-border transactions; Russian researchers claim Chinese APT 31 targeted Russian IT contractors; Australia's weather bureau faces scrutiny over a $62M website upgrade that failed users during severe storms; Google expands AI infrastructure in Taiwan and Singapore; VMware partners with NEC on private cloud modernization in Japan.

See all Cloudflare archives