Not So Fastly

A company you’ve probably never heard of caused half the internet to go dark

Swaths of websites went down on Tuesday morning after an outage at the cloud computing services provider Fastly. Internet users were unable to access major news outlets, e-commerce platforms, and even government websites. Everyone from Amazon to the New York Times to the White House was affected.

At around 6:30 am ET, Fastly said it applied a “fix” to the issue, and many of the websites that went down seemed to be working again as of 9 am ET. Still, the outage highlights how dependent, centralized, and susceptible the infrastructure supporting the internet — especially cloud computing providers that the average user doesn’t directly interact with — actually is. This is at least the third time in less than a year that a problem at a large cloud computing provider has led to countless websites and apps going dark.

Fastly is a content delivery network (CDN), which maintains a network of servers that transfer content quickly from websites to users. The company, which counts Shopify, Stripe, and countless media outlets as customers, promises “lightning fast delivery” and “advanced security.” The nature of such a network also means that problems can quickly spread and affect many of those customers at once. In the case of Tuesday’s incident, Fastly says it “identified a service configuration that triggered disruptions” around the globe. It took about two hours from the time the problem was identified until a fix was implemented.

I encountered this earlier today across several news sites as well as on Bandcamp. What I found most infuriating was that it looks like there wasn’t anything in place to at least — as Cloudflare does — drop to a page with a plain-English explanation that there’s a problem. Instead, I was left with curt error messages that didn’t indicate the source of the problem. I only found out that Fastly was to blame from today’s issue of Dave Pell’s NextDraft newsletter.

The scale of Tuesday’s outage — and frequency of large outages like this one — is what’s really worrisome. Last July, connection issues between two of the data centers operated by Cloudflare ultimately took many sites, including Politico, League of Legends, and Discord, briefly offline. Then, a data-processing problem for Amazon Web Services last November caused problems for sites like the Chicago Tribune, the security camera company Ring, and Glassdoor. The Fastly outage shows the trend continuing, especially as most of the web remains increasingly dependent on cloud providers.

Now I’m wondering if it was a wise move to use Cloudflare to secure and cache my websites.