Cases2 cases

1 to 2 of 2

1. IL-0283 Cloudflare's core proxy returned HTTP 5xx across its network after loading an oversized Bot Management feature file

   Cloudflare2025

   Symptoms

   From about 11:20 UTC on 18 November 2025, Cloudflare's network started failing to deliver core traffic, and Internet users visiting Cloudflare customers' sites got an error page that pointed to a failure inside Cloudflare's network. The volume of HTTP 5xx responses served by the network jumped far above its normal low baseline.

   High error rateService outageHigh latencyResponse quality degradation

   Trigger

   At 11:05 UTC Cloudflare deployed a ClickHouse access control change that made users' existing implicit access to the underlying tables in the r0 database explicit, so the metadata of those tables became visible in system tables such as system.columns. The change was being rolled out gradually across the cluster.

   Root cause

   Configuration change

   Component

   DatabaseCDN and edge network

   Contributing factor

   Unverified assumptionPartial rolloutNo staged rolloutNo configuration testsOverlapping failuresRetry storm

   Detection

   The first automated test detected the issue at 11:31 UTC, three minutes after the first errors on customer HTTP traffic at 11:28. Manual investigation started at 11:32 and the incident call was created at 11:35.

   Remediation

   • Bypassed the core proxy for Workers KV and Access so they fell back to a prior proxy version, which reduced their impact
   • Stopped the automatic creation and deployment of new Bot Management feature files
   • Inserted a known good feature file into the distribution queue and deployed it globally
   • Forced a restart of the core proxy
   • Scaled control plane concurrency to clear the backlog of dashboard logins
   • Restarted the remaining services that had entered a bad state

   Action items

   • Harden the ingestion of Cloudflare-generated configuration files the same way as user-generated input
   • Enable more global kill switches for features
   • Stop core dumps and other error reports from being able to overwhelm system resources
   • Review the failure modes for error conditions across all core proxy modules

   View case
2. IL-0584 Cloud CDN at Osaka served corrupted JavaScript files, and a single-page app failed with a SyntaxError in western Japan

   Monicle2024

   Symptoms

   Some users reported that a Monicle web service could not be used. The service is a single-page app (SPA) served through Google Cloud Load Balancing and Cloud CDN, and in the affected browsers the app stopped with a SyntaxError. Users in eastern Japan were not affected; users in western Japan were.

   Response quality degradation

   Root cause

   Cloud provider outage

   Component

   CDN and edge network

   Contributing factor

   –

   Detection

   Some users reported that they could not use the system, and a war room was opened right away because of the product and the extent of the impact.

   Remediation

   • Purged the Cloud CDN cache after the corruption was seen at KIX (Osaka) but not at NRT (Tokyo)
   • Enabled compression_mode for Cloud CDN content on Cloud Load Balancing, after which the app worked for users again

   Action items

   • Stay aware of which parts of the service are single points of failure, remove them where the cost is justified, and prepare to respond quickly to failures in PaaS and SaaS products

   Lessons

   The author states that there is no specific countermeasure for this incident yet. The lesson drawn is to keep in mind which parts of the service are single points of failure and to be ready to act quickly on failures in PaaS and SaaS products.

   View case

By root cause2