Skip to main content

DevOps & Performance

CDN Outage Analysis: 2026 Impact Report

What broke in Cloudflare's 18 November 2025 and 20 February 2026 outages and AWS's us-east-1 failure of 19 to 20 October 2025, how long each lasted, and which monitoring layers catch them.

Rai AnsarFounder of Visual Sentinel13 min read

CDN Outage Analysis: 2026 Impact Report

The November 2025 Cloudflare outage wasn't just another service disruption. It was a stark reminder of how deeply our digital infrastructure depends on a handful of CDN providers. In my experience monitoring production systems, I've watched teams scramble as their carefully architected applications became unreachable not because of their own failures, but because their CDN went dark.

Three incidents from the past year show the pattern: Cloudflare's outage on 18 November 2025, AWS's us-east-1 failure on 19 to 20 October 2025, and Cloudflare's BYOIP outage on 20 February 2026. Each had a different cause, and each looks different from a monitoring check. This report covers what happened, how long it lasted, and what to monitor.

Major CDN Outages of 2025 and 2026: What Broke and for How Long

Each incident below is taken from the provider's own postmortem, which is linked in the section.

Cloudflare's Global Outage (18 November 2025)

This cdn outage analysis starts with the biggest one. According to Cloudflare's postmortem, the network began failing to deliver core traffic at about 11:20 UTC on 18 November 2025. Core traffic was largely flowing normally by 14:30 UTC, and all systems were back to normal at 17:06 UTC. Visitors to sites behind Cloudflare saw HTTP 5xx errors. Turnstile, Workers KV and Access were also affected, and most people could not log in to the Cloudflare dashboard.

The cause was an internal change, not an attack and not a traffic spike. A database permissions change made a query return duplicate rows, and those rows went into the feature file used by Cloudflare's Bot Management. The file doubled in size and exceeded a limit in the proxy software (200 features, against about 60 in use), so the proxy crashed.

The file was regenerated every five minutes while the database change was still rolling out, so the network flipped between working and failing. That is what a DDoS attack can look like, and Cloudflare says its team first wrongly suspected one. The outage cost time partly because the symptoms pointed at the wrong cause.

AWS Infrastructure Failures (19 to 20 October 2025)

AWS is not a CDN provider, but this incident matters to anyone who serves a site through a CDN, because origins, APIs and sign-in often sit in one region. In AWS's post-event summary, the event ran from 11:48 PM PDT on 19 October to 2:20 PM PDT on 20 October 2025, about 14.5 hours, in the us-east-1 (N. Virginia) region.

The root cause was a latent race condition in the automation that manages DNS for DynamoDB. It left the regional DynamoDB endpoint with an empty DNS record, so for about 2 hours 37 minutes nothing could resolve an address for it. That broke DynamoDB first, then services that depend on it: EC2 instance launches, Network Load Balancer health checks, Lambda, STS and AWS console sign-in. AWS disabled the affected DNS automation worldwide while it fixed the race condition.

I've seen teams assume a provider of this scale is immune to this kind of failure. This incident was limited to one region and cascaded through the services that depend on DynamoDB. That is still enough to take down a lot of sites.

Cloudflare's BYOIP Outage (20 February 2026)

On 20 February 2026, a bug in a new automated cleanup task at Cloudflare caused about 1,100 of 4,306 BYOIP (Bring Your Own IP) prefixes to be withdrawn from the internet over BGP. The task called an API with an empty pending_delete value, which the API read as "all prefixes". Cloudflare's postmortem says the outage began at 17:48 UTC in the introduction (the timeline marks impact start at 17:56), and the last affected prefixes were restored at 23:03 UTC. Core CDN, Spectrum, Magic Transit and Dedicated Egress were affected, and Cloudflare says it was not an attack.

Geographic Impact Distribution

Provider failures have a location. AWS's event was in us-east-1, so a team with its database or login service in that region felt it even if the site itself ran elsewhere. Cloudflare's BYOIP outage affected customers who bring their own IP prefixes to Cloudflare. Check where your CDN, origin, DNS and sign-in dependencies sit, and set up monitoring points that would notice a failure in each of those places.

Root Causes: What's Really Breaking CDN Infrastructure

Analyzing the underlying causes of these CDN failures reveals systemic issues that extend beyond simple hardware problems.

DNS Automation Failures

DNS automation is a weak point in modern infrastructure. In the AWS October incident, two DNS Enactors that update DNS records raced each other, and the result was a regional endpoint with no addresses at all. The automation could not repair that state, so engineers had to fix it by hand. When DNS automation fails, it doesn't just break one service. Everything that looks up that name fails with it.

I've observed that teams often implement DNS automation without sufficient validation layers. The complexity of modern DNS configurations, especially with features like traffic steering and geographic routing, creates multiple failure points that can trigger widespread outages.

Misdiagnosis: Attack or Internal Failure?

A burst of errors that comes and goes looks like an attack. In the November incident, Cloudflare's own team first suspected a hyper-scale DDoS attack, because the network kept flipping between failing and recovering. The real cause was a bad file being regenerated every five minutes.

In my experience monitoring high-traffic applications, the first question is whether the failure is yours, your provider's or an attacker's. Check the provider's status page, then compare errors across regions and across customers. If the same 5xx errors appear on unrelated sites behind the same CDN, look at the provider first. Don't assume every error response is an attack.

Configuration Error Patterns

The February 2026 BYOIP outage and the November 2025 outage share a pattern: an internal change behaved differently from what its authors expected. In February it was a cleanup task that treated an empty parameter as "everything". In November it was a database permissions change that nobody expected to change a query's output.

These errors are dangerous because they hit many services at once. When a core networking or data file fails, CDN, security services and traffic routing can all be affected together.

Cascading Effects: How CDN Failures Impact Website Monitoring Metrics

CDN outages create unique challenges for monitoring systems because they affect multiple infrastructure layers simultaneously. Understanding these cascading effects is crucial for effective cdn outage analysis.

Unreachable Even When the Origin Is Healthy

When a CDN fails, visitors can't reach your site even though your origin server is fine. I've watched monitoring dashboards light up with errors during CDN outages while the origin's own health checks stayed green. In the BYOIP outage, Cloudflare withdrew the routes for the affected IP prefixes, so connections to them timed out or failed. Nothing was wrong with the customers' servers.

A check that only runs from one location, or only against the origin, can miss this. Run checks against the public hostname that visitors use, from more than one region, so that a failure in the CDN path shows up.

DNS Resolution Timeouts

DNS failures during provider outages create some of the most confusing monitoring scenarios. In the AWS October incident, the DynamoDB endpoint had no IP addresses for about 2.5 hours, so anything that depended on that name failed at the lookup step, before any connection was attempted. That is hard to tell apart from a local network problem unless your monitoring checks DNS separately from uptime.

Cached answers prolong the pain. AWS reports that after the records were restored, customers recovered fully only as cached DNS records expired. Allow for your own record TTLs when you judge how long recovery will take.

Performance Degradation Patterns

CDN failures can manifest as severe performance degradation before complete failure occurs.

Intermittent connectivity and increased response times can precede complete failure. This degradation can confuse monitoring systems that use simple binary up/down checks.

I've learned to configure monitoring thresholds that detect performance degradation patterns rather than just complete failures. This approach provides earlier warning of CDN-related issues.

Multi-Layer Monitoring Strategies for CDN Resilience

Effective CDN monitoring requires a comprehensive approach that accounts for the complex interdependencies in modern web infrastructure.

Proactive CDN Health Checks

Monitoring CDN health goes beyond simple endpoint connectivity. I recommend implementing checks that verify actual content delivery, not just network reachability.

Key monitoring layers include:

  1. Uptime monitoring from multiple geographic locations
  2. Content delivery verification to ensure cached resources are accessible
  3. Response time tracking to detect performance degradation before complete failure
  4. Error rate monitoring to catch increasing failure rates early

Tools like Pingdom, UptimeRobot, and Visual Sentinel's uptime monitoring can provide these capabilities, but the key is ensuring geographic distribution of monitoring points.

Geographic Distribution Testing

The geographic nature of CDN outages makes distributed monitoring essential. A CDN failure in Europe might not immediately affect North American users, but it will impact your European customer base significantly.

I've seen teams rely on single-location monitoring only to discover their CDN was failing in specific regions while appearing healthy from their primary monitoring location. This geographic blindness can mask serious customer impact.

Effective geographic monitoring requires:

  • Monitoring points in all major regions where you have users
  • Region-specific alerting to avoid false positives
  • Correlation analysis to distinguish between local and global CDN issues

Configuration Validation

Given that automation and configuration errors caused major outages like the AWS DNS incident, proactive configuration monitoring becomes critical. This includes validating DNS settings, SSL certificate configurations, and CDN routing rules before they're deployed.

I recommend implementing automated configuration validation that runs continuously, not just during deployment. CDN configurations can drift over time, and periodic validation helps catch issues before they cause outages.

What These Outages Show for 2026

Three incidents are too few to call a trend, but each one shows how a single fault at a shared provider reaches many sites.

Outages Do Not Wait for a Quiet Season

The three incidents above started on a Tuesday morning in UTC, on a Sunday night in California, and on a Friday evening in UTC. None waited for a maintenance window or a quiet period, so a monitoring plan that assumes quiet periods will leave gaps. Alerts have to reach someone at any hour.

Public Cloud Risk Concentration

Shared public cloud infrastructure creates concentration risk as services migrate to the same providers.

The concentration effect means that single-provider failures can impact dependent services simultaneously. The AWS October incident shows how shared dependencies amplify the impact of an outage.

Emerging Vulnerability Points

The February 2026 Cloudflare incident came from BYOIP, the feature that lets customers bring their own IP prefixes, and a new automated cleanup task. As CDN services add features, they add new ways to fail.

I expect to see more configuration-related outages as teams adopt advanced CDN features without fully understanding their failure modes. The trade-off between functionality and reliability becomes more pronounced with each new feature.

Building CDN Outage Detection into Your Monitoring Stack

Creating effective CDN outage detection requires integrating multiple monitoring approaches into a coherent strategy.

Essential Monitoring Layers

A comprehensive CDN monitoring strategy must address multiple failure modes simultaneously:

Monitoring TypePurposeKey Metrics
UptimeDetect complete failuresResponse codes, connectivity
DNS MonitoringCatch resolution issuesQuery response times, propagation
SSL MonitoringVerify certificate deliveryCertificate validity, chain completion
PerformanceDetect degradationResponse times, throughput
ContentVerify delivery integrityContent checksums, cache status

Each layer provides different insights into CDN health. Uptime monitoring might show green while DNS resolution is failing, or SSL certificates might be invalid while basic connectivity appears normal.

Alert Configuration Best Practices

CDN outage alerts require careful tuning to avoid both false positives and missed incidents. Based on my experience, I recommend:

  1. Threshold-based alerting that accounts for normal CDN performance variation
  2. Geographic correlation to distinguish between local and global issues
  3. Escalation patterns that account for CDN provider response times
  4. Integration with CDN provider status pages to correlate internal monitoring with provider communications

The key is balancing sensitivity with specificity. CDN performance can be highly variable, so alerts need to account for normal fluctuations while still catching real problems early.

Failover Automation

Automated failover to secondary CDN providers can minimize the impact of primary CDN outages. However, implementing effective failover requires careful consideration of DNS TTL values, traffic routing logic, and performance implications.

I've seen teams implement failover systems that work perfectly in testing but fail during real outages due to DNS caching or traffic routing complexities. The key is testing failover scenarios regularly and ensuring your monitoring can detect when failover has occurred.

The Cost of CDN Dependencies

The financial impact of CDN outages extends far beyond the immediate service disruption. Say an outage costs a business $2,000 an hour (an example figure, not a benchmark). Cloudflare's November 2025 incident ran from 11:20 to 17:06 UTC, which is 5 hours 46 minutes, with core traffic back at 14:30. At that rate, the roughly three hours until core traffic recovered would cost $6,000, and the full 5 hours 46 minutes about $11,500, for that one business. The same outage hits other dependent sites at once.

In my experience, the hidden costs often exceed the obvious ones. While teams focus on lost revenue during outages, the engineering time spent on incident response, customer communication, and post-incident analysis can be enormous.

The concentration risk in modern CDN infrastructure means these costs are often shared across many organizations simultaneously. When Cloudflare or AWS experiences major outages, entire sectors of the internet economy feel the impact.

Building Resilience for the Future

The cdn outage analysis from 2025-2026 provides clear lessons for building more resilient web infrastructure. What I would do on a Monday: monitor your site from outside your CDN, so that you see what visitors see. Add DNS and SSL checks, plus a content check that proves real pages are being served. Write down who gets alerted and what the failover step is, then test the failover.

All three incidents started on the provider's side, so a monitor that sits outside your CDN and checks DNS, certificates and page content from more than one place is what tells you quickly that the fault is the provider's and not yours.

Visual Sentinel checks your site from Europe and the US. Your first check needs no card, and the 10-day free trial (no card with a work email) lets you add uptime, DNS and SSL monitoring for the public hostname.

To hear when your own site goes down, see how uptime alerts work in Visual Sentinel.

cdn-outagewebsite-monitoringinfrastructure-failuresperformance-monitoringdns-monitoring

Check your site while you read

Run a free check for regional reachability, response time, redirect handling, TLS evidence, and DNS resolution. No account is required.

Check a site free

Common questions

What was the most significant CDN outage in 2026?

The largest in this report is Cloudflare's outage on 18 November 2025, which ran from about 11:20 to 17:06 UTC, caused by an oversized Bot Management feature file after a database permissions change. The 2026 incident covered here is Cloudflare's BYOIP outage on 20 February 2026.

How do CDN outages affect website monitoring metrics?

CDN and provider outages can hit several monitoring layers at once: uptime checks return 5xx errors or time out, DNS lookups fail (as with the AWS DynamoDB endpoint), and content checks fail because pages are not served. Checking each layer separately tells you which one broke.

What monitoring strategies help detect CDN failures early?

Effective CDN monitoring requires multi-layer detection including uptime monitoring from multiple geographic locations, DNS resolution testing, SSL certificate delivery verification, and content change detection to ensure CDN cache integrity.

How much do CDN outages typically cost businesses?

It depends on the business. As an example, a site that loses $2,000 an hour would lose about $6,000 over the three hours it took Cloudflare's November 2025 outage to recover core traffic, and about $11,500 over the full 5 hours 46 minutes. The same outage hits many dependent sites at once.

Are CDN outages becoming more frequent in 2026?

Three incidents are too few to measure a trend. What they show is concentration risk: one fault at a shared provider can reach many sites at once.

What are the main causes of CDN outages?

In the incidents covered here: a database permissions change that produced an oversized configuration file (Cloudflare, November 2025), a race condition in DNS automation (AWS, October 2025) and a bug in a cleanup task (Cloudflare, February 2026). Hardware failures and traffic attacks also cause outages, but none of these three was caused by either.

Related

  • Worst Infrastructure Monitoring Mistakes

    infrastructure monitoring

    Worst Infrastructure Monitoring Mistakes

    Overlooked monitoring layers can leave infrastructure issues unnoticed. This guide explores uptime, SSL, DNS, visual and content checks that help sysadmins investigate website problems.

    Rai Ansar7 min read

  • CDN Outage Analysis: Fastly & AWS Cases

    industry news

    CDN Outage Analysis: Fastly & AWS Cases

    Fastly's June 2021 outage and AWS US-EAST-1 DNS failures in Q4 2025 show how infrastructure dependencies can disrupt websites. This analysis explores the causes, operational impacts and role of monitoring in incident response.

    Rai Ansar10 min read

  • Ransomware Detection via Site Monitoring

    security monitoring

    Ransomware Detection via Site Monitoring

    Which web-layer signals can point to a ransomware compromise (injected links, redirects, ransom wording, changed screenshots), what to do when you see them, and which server-side tools cover the rest.

    Rai Ansar10 min read

Stop guessing whether your site still works.

Add your site and its uptime, SSL certificate and screenshots are checked for 3 days. Then try a paid plan free for 10 days (no card with a work email) or subscribe, from $15 USD a month.

Start 10-day trialSee pricing