What Actually Happens When Cloudflare Goes Down (DNS, CDN, Edge Explained)

On November 18, 2025, Cloudflare suffered a global outage that broke roughly one-third of the world's 10,000 most popular websites. X (formerly Twitter), ChatGPT, Spotify, Zoom, Coinbase, and even outage-tracking sites like Downdetector went dark for hours. For most users, it looked like "the internet broke."

Less than three weeks later, on December 5, 2025, it happened again. This time, 28% of Cloudflare's traffic failed for 25 minutes.

What actually breaks when Cloudflare goes down? And why does so much of the internet depend on a single company?

## The Three Services That Break

Cloudflare isn't just one thing. It's three critical infrastructure layers stacked on top of each other: DNS, CDN, and Edge Computing. When Cloudflare has an outage, all three can fail independently or together.

### DNS (Domain Name System)

DNS is the internet's phonebook. When you type example.com into your browser, DNS translates that human-readable domain into an IP address like 104.21.1.1 that computers actually understand.

Cloudflare runs one of the world's fastest DNS resolvers at 1.1.1.1. Millions of sites also use Cloudflare as their authoritative DNS provider — meaning Cloudflare is responsible for answering "what IP address does this domain point to?"

When Cloudflare's DNS goes down:

• Your browser can't resolve domain names
• Even if the origin server is healthy, users can't reach it
• The entire site appears offline, even though it's technically running

Real-world impact: During the September 2024 outage, Cloudflare accidentally stopped announcing fifteen IPv4 prefixes during routine maintenance. For one hour, 1,661 Business plan websites became unreachable. Traffic dropped by 10 Gbps instantly. Browsers showed DNS resolution errors even though the actual websites were fine.

### CDN (Content Delivery Network)

A CDN caches static assets like images, CSS, and JavaScript files across hundreds of data centers worldwide. Instead of fetching a file from a single origin server in Virginia, a user in Tokyo gets it from a nearby Cloudflare edge server in milliseconds.

Cloudflare's CDN sits in front of millions of websites. When users request a page, they hit Cloudflare first — not the origin server.

When Cloudflare's CDN goes down:

• Cached content becomes unavailable
• Traffic can't reach the origin server (Cloudflare acts as a reverse proxy)
• Even dynamic requests fail because they never make it past the CDN layer
• SSL/TLS certificates managed by Cloudflare can't be validated

Real-world impact: During the June 20, 2024 incident, two independent failures caused 114 minutes of degraded performance. At the peak, 2.1% of HTTP requests received generic error pages. Sites proxied through Cloudflare returned 502 Bad Gateway errors. The 99th percentile Time To First Byte latency tripled, even though origin servers were responding normally.

### Edge Computing (Workers, Pages, R2)

Cloudflare Workers let developers run JavaScript code at the edge — before requests even reach the origin server. This powers serverless functions, A/B testing, authentication, and entire applications.

When Cloudflare's edge platform goes down:

• Serverless functions stop executing
• Authentication middleware fails
• API endpoints hosted on Workers return errors
• Sites built entirely on Cloudflare Pages go offline

Real-world impact: On February 6, 2025, Cloudflare's R2 object storage service failed due to credential rotation errors. Sites relying on R2 for media storage couldn't serve images or videos. The outage lasted over an hour. On March 21, 2025, another R2 failure caused total write failures for 67 minutes, disrupting file uploads across thousands of applications.

## Why So Many Sites Use Cloudflare

Cloudflare isn't just popular — it's become critical infrastructure. Here's why:

It's free for small sites. Cloudflare's free tier includes DNS, CDN, DDoS protection, and SSL certificates. For indie developers and startups, it's a no-brainer.

It's fast. Cloudflare operates 330+ data centers worldwide. A user in Singapore gets content from Singapore, not from a server in Iowa.

It stops attacks. Cloudflare blocks 150+ billion cyber threats per day. Without it, many sites would be vulnerable to DDoS attacks that overwhelm servers with fake traffic.

It's invisible. Most developers set it up once and forget about it. DNS records point to Cloudflare, traffic flows through their network, and everything just works — until it doesn't.

## The Single Point of Failure Problem

The 2025 outages exposed a harsh reality: Cloudflare has become a single point of failure for a significant chunk of the internet.

Here's the typical architecture:

User → Cloudflare DNS → Cloudflare CDN → Cloudflare Workers → Origin Server

Every layer depends on Cloudflare. If any one piece fails, the entire chain breaks.

The November 18, 2025 incident was triggered by a bug in Cloudflare's Bot Management feature. The bug caused a configuration file to grow beyond its expected size, crashing the software system handling traffic across multiple services. The outage lasted several hours and affected major platforms including X, ChatGPT, and Spotify.

The December 5, 2025 incident happened while Cloudflare was protecting customers from a React framework vulnerability. A configuration change meant to turn off an internal testing tool inadvertently caused proxy errors under certain conditions. Within seconds, the change propagated to the entire fleet of servers globally, taking down 28% of traffic for 25 minutes.

Both failures followed the same pattern: a seemingly routine configuration change, deployed instantly across hundreds of data centers worldwide, triggered cascading failures. Even though origin servers were healthy, users couldn't reach them because Cloudflare's infrastructure layer failed.

After these back-to-back outages, Cloudflare declared "Code Orange: Fail Small" — a company-wide initiative prioritized above all other work to prevent global outages from single configuration changes.

## What Developers Can Do

If you run a site on Cloudflare, here are practical steps to reduce risk:

Use multiple DNS providers. Configure secondary nameservers with Route53 or Google Cloud DNS. If Cloudflare's DNS fails, browsers will fall back to the secondary provider.

Keep a bypass ready. Store your origin server's IP address and test direct connections regularly. During an outage, you can manually update DNS records to point directly to your server.

Monitor from outside Cloudflare. Use uptime monitors like UptimeRobot or Pingdom that check your site from multiple global locations. If Cloudflare is down but your origin is up, you'll know immediately.

Cache critical assets elsewhere. Store a copy of essential JavaScript and CSS files in a separate CDN like Fastly or AWS CloudFront. If Cloudflare fails, your site can still load core functionality.

Don't put authentication behind Workers. If Cloudflare Workers go down, users locked behind edge-based authentication can't access your site at all. Keep authentication on your origin server when possible.

Test your disaster recovery plan. Simulate a Cloudflare outage by temporarily removing their DNS records and pointing directly to your origin. Does your site still work? Can users still log in?

## The Bigger Picture

Cloudflare's success has created a paradox. The same company that protects millions of sites from outages has become a potential source of massive outages itself.

The internet was designed to be decentralized — no single point of failure. But in practice, economic incentives push everyone toward a handful of providers. Cloudflare, AWS, and Google collectively power most of the web.

According to Cisco's network monitoring data, major outages increased from 10 in 2022 to 23 in 2024, with 12 recorded in just the first half of 2025. Common patterns include configuration changes that cascade across systems, silent failures in seemingly healthy systems, and changes that spread failures rather than contain them.

The November and December 2025 Cloudflare outages are wake-up calls. They remind us that the internet's resilience depends on intentional architectural choices, not just default configurations. As one Cloudflare engineer put it after the incidents: "Work is already underway to make sure it does not happen again, but I know it caused real pain today."

The question isn't if another major outage will happen. It's when — and whether your site will survive it.

---

Thanks for reading! If you found this helpful, connect with me on LinkedIn or check out my work on GitHub.