05 minute read

Cloudflare went south

One company’s outage rippled across the web. From chatbots to social feeds, the tools we rely on every day suddenly became unreachable.

Last month, Amazon brought most of the internet to its feet when a major outage occurred in the very well known US-EAST-1 region. Not even a month later, it’s Cloudflare’s time to take the stage.

By now, there are plenty of memes around this subject, some argue it was an intern messing with Cloudflare’s lava lamp wall, and others go for the usual ‘oopsie, wrong cable’ parody.

There’s no way around it

The reality is that both Amazon’s and Cloudflare’s infrastructures are so deeply connected to so many websites and services that when one of these giants coughs, the whole internet tumbles.

Even when you think you are going to escape AWS and give your cash to some local company, chances are they are just renting AWS themselves and allow you to take a portion of their storage for a significantly higher fee than what you would have payed Amazon in the first place.

Something’s wrong

I’m on a relatively new computer—I wrote about this before—so it’s all still very dubious to me when it comes to replicating my usual setup on a new machine.

For instance, I use Cloudflare’s Zero Trust (CFZT) on all my devices. This allows me to easily block ads on any device I may be using at the DNS level. Which basically means I don’t need to worry about a new app or a new OS or even a new device. It happens way before those are even relevant.

I believe at the time of the outage I was trying to access ChatGPT to help me config something in Linux. And I was displayed a message along the lines of “Please unblock challenges.cloudflare.com to proceed”.

So I thought that perhaps my ad blocking script in CFZT had been updated overnight with this URL and ChatGPT would need access to it to function properly, so it refused to work. Ok no worries, I disconnected from CFZT and tried again. Same issue.

At this point I’m wondering whether or not I had this ad blocking script also set up in my router—something that I actually wondered before implementing it but eventually decided against because that approach would only cover for the time I was at home. And it wasn’t there either.

Let’s move on

Well, I had other stuff to do so I paused this Linux config I was after and decided to move forward with some other task.

This other task was pushing an update to https://predifix.pt. I just merged with master and went ahead to the next task. A few minutes afterwards I go back to check the website was actually updated. And to my surprise, it wasn’t. So I head back to GitHub to see what went wrong. This website is deployed with Cloudflare Pages, and as we all know by now, that was also down. So when I tried to investigate the build logs—which point to Cloudflare—I also had no luck.

But at the time I didn’t know what it was, so I wanted to check if something was wrong with Cloudflare. Just like any human being I go to Down detector to see if anyone else was reporting issues.

Guess what… Down detector was itself down because it also depends on Cloudflare. Next best thing? Twitter—yes, I have a problem referring to it as X. Can you guess what happened? Right! It was also down!

See how the whole internet gets screwed when one of the giants falls?

So it’s not just me

So ChatGPT isn’t working, Downdetector is down, Twitter isn’t working either, Cloudflare Pages aren’t deploying… I have to assume it’s not just me at this point.

Truthfully told, I thought it had been Amazon again! But apparently this time it was a bit closer to my workflow.

All of my websites either depend on Cloudflare for DNS, or are actually built by Cloudflare Pages themselves. So for me it was far more serious than the Amazon’s incident.

Who knew a few servers in one company could make the whole internet act like it forgot how to work? For a few hours, we were all gently reminded that the web isn’t magic—it’s fragile, and occasionally hilarious.

It’s easy to forget how much of what we rely on every day lives in someone else’s hands. An outage like this nudges us to pause, notice the infrastructure behind the scenes, and appreciate the invisible threads keeping us connected.

Photo of Pedro