Cloudflare Outage Hits Internet Services for Second Time in a Week

A short but widespread outage at the internet service company Cloudflare disrupted access to many of the web’s most popular sites, like Facebook, Google and Amazon, on Tuesday morning — providing a fresh reminder of how many of the tech world’s basic functions are dependent on a small number of behind-the-scenes companies. The company admitted to a network problem and said it already implemented a fix minutes after the outage, with services gradually coming back up shortly thereafter.

What happened during Cloudflare’s brief global outage

Cloudflare said that it first experienced the outage at 8:56 a.m. UTC and that it resolved the issue at 9:12 a.m., in an incident less than 20 minutes long — but through which its effects were felt across the internet’s infrastructure far from its network edges. Although the company did not immediately specify the root cause, it lines up with a pattern we’ve seen many times in high-scale infrastructure: some configuration or routing issue causes mass timeouts; clients follow this with — you guessed it — mass retries and dependency chains broaden the blast radius.

Since Cloudflare is a content delivery network (CDN), reverse proxy, and security component for a wide range of websites, these disruptions are felt in multiple layers including DNS resolution, TLS termination, or request routing. Even a momentary hiccup in the control plane can starve user sessions and API calls for any of tens or hundreds of unrelated brands.

Which websites and users were affected by the outage

There were reports of people being unable to access LinkedIn, Epic Games, Shopify, Canva, Crunchyroll and Zoom, among other sites. Ironically, outage-tracking posts had their own problems as anecdotal reports poured in. Symptoms differed from site to site: some sites failed to load at all, others returned gateway errors, and a few apps hung during login or checkout flows as third-party scripts and APIs timed out.

Enterprises felt it too. Background jobs, webhooks, and payment callbacks can get stuck when an edge proxy or DNS service runs into trouble. In e-commerce, 5xx errors for a few minutes, even after the network has stabilized, can mean abandoned carts and customer support surges.

Why an edge provider’s outages can spread so widely

Cloudflare sits in the request path to a huge chunk of the web, speeding up content, stopping malicious traffic and protecting origins. That centrality is an advantage for performance and security, but a liability in a crisis: if one of the providers’ edges pop or backbone routing goes wonky for some reason, that blip can traverse continents in seconds.

Network analysts working at companies like Kentik and ThousandEyes have demonstrated how BGP changes, DNS failures and configuration rollouts can spread in subtle ways. They may have never seen the scalability-beating “whoops, I broadcast a withdrawal to my transit with thousands of prepends” incidents experienced by their peers in the industry, such as 2021’s global outage associated with a Fastly configuration bug or Cloudflare’s 2022 event blamed on a backbone change — where even just one misplaced step of automation or routing policy creates outsized amounts of pain.

Cloudflare’s response and recovery timeline for today

How fast that incident was fixed suggests operators quickly isolated the offending change and rolled it back, much faster than the hour-long recovery of a recent prior Cloudflare disruption of multiple hours. In high reliability networks, a reduced mean time to recover from incidents references strong observability and well-rehearsed incident playbooks: quick identification, traffic shedding, selective reversal of changes and gradual reintroduction of capacity.

But even brief outages can have outsized consequences. DNS caches may contain stale answers, clients may be overly aggressive in backoff or retry, and microservices can experience cascading failures when they lack circuit breakers. A long tail of intermittent errors is often observed, as ecosystems re-synchronize.

What this outage means for builders and everyday users

This is another push in the direction of resilience patterns for engineering teams: multi-CDN, dual-authoritative-DNS providers, canarying and progressive conf rollouts, client-side timeouts which fail-safe. Strategies such as stale-while-revalidate, origin shields, and graceful degradation can keep critical paths alive while the edge blinks.

For users, the lesson is simple: when a big edge provider falls down, wait a few minutes before trying to log in or pay in order to avoid making extra work for back-end servers. For organizations, post-incident reviews need to put a dollar figure on just 10–20 minutes of downtime, and compare it against an investment in redundancy.

The broader infrastructure picture and shared risks

The resilience of the internet is derived more and more from a small number of backbone carriers and edge platforms, including Cloudflare, Akamai, Fastly and major cloud providers. Concentration results in efficiency and strong circular defenses against attacks, but it also generates systemic risk when an operator fails.

These incidents are often recorded in near real time by impartial observers like Cloudflare Radar and the industry telemetry from companies such as Apmel, Catchpoint and NetBlocks — demonstrating just how intertwined the web’s building blocks have become.

The rapid recovery we saw today is heartening, but the mistake underlines a larger lesson: redundancy isn’t a luxury — it’s table stakes for a connected economy.