Cloudflare outage disrupts LinkedIn, Zoom and more

A short, widespread outage at the web service company Cloudflare recently affected several major websites, including OkCupid, Uber, and Have I Been Pwned, disabling them temporarily across many parts of the internet. The company confirmed the incident, pushed through a fix in about half an hour, and said it was monitoring recovery as traffic and user reporting began to return after nearly an hour.

What happened during the brief Cloudflare outage

Cloudflare’s status page flagged a network problem that caused sporadic failures for sites and apps using its edge network. Symptoms varied by service: some users encountered login loops or messages that said “connection failed,” others had assets like images and scripts time out. Observation site Downdetector registered simultaneous spikes in outage reports across various platforms that quickly receded after Cloudflare pushed a fix.

Curiously, Cloudflare also disclosed a separate issue impacting its Workers AI platform during the same timeframe.

Early indications were that those problems were distinct from the broader customer-facing outages, but in both cases they underlined what a house of cards large-scale, globally distributed systems can be — sensitive to even small configuration and platform errors.

Scope and affected services across major platforms

Most attention was focused on LinkedIn and Zoom because of their size and immediate impact on work communications, but consumer and creator platforms like Canva and Crunchyroll are also on the list, along with on-demand services such as DoorDash and home security provider Ring. Many functions continued to work at least in part, especially for content already cached at the edge or in app sessions for which no fresh authentication was necessary.

The pattern is similar to a standard CDN and reverse-proxy event: DNS lookups, TLS handshakes, API calls — all of these could impact you in different ways depending upon the network route, geographic location, and caching state. So two users both sitting in the same city can end up with completely different results within the same event.

Why a Cloudflare blip can disrupt the wider web

Front and center, Cloudflare serves as a content delivery network, DNS provider, and security layer to millions of sites. Cloudflare adoption is estimated at over 20% of the top one million domains using the W3Techs web technologies survey engine, and Cloudflare’s network is moving more than 27 million HTTP requests per second for URL shorteners and just about everyone else. Currently, it operates a high-performance Anycast cloud around the world. When a control-plane or config error is pushed to that footprint, even if it’s only live briefly, the holler can be loud and the lights bright.

The architecture also helps explain why outages can be lumpy. The cache can load cached pages while API endpoints fail; static assets may be rendered while sign-ins reject tokens; mobile clients with long-lived connections can hang waiting for responses while new sessions are blocked. Observability companies like ThousandEyes and Kentik often compile this pattern of “partial breakage” when the internet erupts in failure.

Recent history raises the stakes for Cloudflare outages

The disturbance is an aftershock of a bigger Cloudflare event several weeks earlier, its most serious since 2019. In this latest event, a “feature file” storing data ballooned in size and spread across the network, overwhelming machines and causing significant disruptions that affected services like ChatGPT, Uber, and X. The process of containing it took far less time — with recovery already visible within minutes of deploying a fix.

What Cloudflare has said so far about the outage

Cloudflare said a fix was installed and it was observing results but did not immediately identify a cause. The company also posted support status updates for scheduled work taking place in data centers in Chicago and Detroit, though it’s unclear whether maintenance played any role in the outage. There should eventually be a more formal postmortem, because that is Cloudflare practice for significant incidents.

What consumers and users can do during CDN outages

Organizations with heavy reliance on one particular edge provider should consider revisiting their resiliency plans. Simple workarounds include dual-provider authoritative DNS, multi-CDN with health-check routing, origin failover, and short TTLs for quick configuration rollback. Clients can also limit these kinds of failures (cascading collapses) with a high percentage of synthetic monitoring and explicit “circuit breaker” logic in clients.

The best thing for average users to do during a CDN outage is typically nothing: wait for service restoration and resist the urge to reset passwords multiple times or reinstall apps. Consulting provider status pages and official support channels can help confirm if the problem is local or more widespread. In this particular rapid recovery case, sessions in the cache recover spontaneously when system state (i.e., upstream connectivity) stabilizes.

The bottom line on this brief but disruptive outage

It was a sudden wake-up call to how much the internet depends on a handful of key intermediaries. Even a 30-minute hitch at a network as central as Cloudflare’s can send ripples through work, commerce, and entertainment. The quick fix is tunnel-light, but two successive outages will only increase calls for layered redundancy — and clear-eyed postmortems that turn lessons into longer-term stability.