---
title: "Small Mistakes That Took Down Big Parts of the Internet — Cloud Changelog"
description: "A typo at Amazon, a setting at Fastly, a cleanup at Rogers. Six true stories of small slips at single companies that knocked out services for millions."
url: "https://cloudchangelog.com/articles/small-mistakes-big-internet-outages"
updated: "2026-11-19"
---

# Small Mistakes That Took Down Big Parts of the Internet

Six true stories of one small change at one company that left millions of people staring at error pages.

November 19, 2026

On the morning of June 8, 2021, people around the world opened the BBC, The New York Times and the Financial Times and got an error page instead of the news. So did visitors to Reddit, Twitch, the White House and the British government's main website.[\[1\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-1) None of those organizations had done anything wrong. They all happened to pay the same company, Fastly, to deliver their pages quickly, and Fastly had a bad hour.

"The entire Internet has become dangerously geared on just a few players," Gaz Jones, the technical director of a digital agency, told Fortune that day.[\[1\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-1) The stories below show what that looks like up close. Each one starts with a single person, a single command or a single small change, at one company most people have never thought about.

## Amazon's typo, 2017

On February 28, 2017, at 9:37 in the morning Pacific time, a member of Amazon's storage team was trying to fix a billing system that had been running slowly. The plan was to take a few servers offline with a standard command. One part of the command was typed wrong, and far more servers went offline than intended.[\[2\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-2)

Those servers kept track of where everything in Amazon's storage service lived. Without them, the service stopped answering for hours in its northern Virginia region. Slack, Trello, Quora, Medium, Imgur, Twitch and Expedia all broke. Some Nest security cameras stopped recording for hours.[\[3\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-3)

The best detail came from Amazon's own status page, the one meant to tell customers when something is wrong. It kept showing happy green icons through the worst of it. The red warning icons were stored on the same service that had just gone down.[\[3\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-3) Amazon later admitted it could not update the page for about two hours, and posted news on Twitter instead.[\[2\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-2)

Storage was back to normal a little before 2 p.m., about four hours after the typo. Amazon apologized and changed its tool so that it could no longer remove so many servers so quickly.[\[2\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-2)

## Fastly's hidden bug, 2021

The Fastly outage that opened this article had started weeks earlier. On May 12, 2021, Fastly rolled out a software update that contained a bug nobody noticed. It sat there quietly for almost a month.[\[4\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-4)

Early on June 8, one Fastly customer changed a setting on their own account. It was a normal, allowed change. But it happened to be exactly the kind of change that woke the bug up, and within moments 85 percent of Fastly's network was sending back errors.[\[4\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-4)

That is how a single customer's settings ended up taking down newspapers, Reddit and the British government's website.[\[1\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-1) Fastly's own account said the change was perfectly allowed, and it did not name the customer.[\[4\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-4)

Fastly's monitoring spotted the trouble within a minute. Within 49 minutes, most of its network was working again.[\[4\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-4) Nick Rockwell, a senior Fastly executive, wrote up what had happened and apologized to Fastly's customers and to everyone who relies on them.[\[4\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-4) For most people the whole thing lasted about an hour.[\[1\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-1)

## Facebook locks itself out, 2021

Around noon Eastern time on October 4, 2021, Facebook, Instagram and WhatsApp disappeared. They stayed gone for about six hours, the longest stretch of downtime Facebook had suffered since 2008. Around 3 billion people use those apps.[\[5\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-5)

The cause was routine maintenance. Someone ran a command meant to check how much spare capacity Facebook's private network had. Instead, the command cut every Facebook data center off from the network. A safety tool was supposed to stop commands like that one, but a bug in the tool let it through.[\[6\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-6)

Once the data centers went dark, the rest of the internet could no longer find Facebook at all. Worse, Facebook's own engineers could not reach their systems remotely, because that access ran over the same broken network. The company had to send people to the data centers in person. Even that took time, because those buildings are designed to be very hard to get into.[\[6\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-6)

There were reports that employees at Facebook's California headquarters had trouble getting into their own buildings that day.[\[7\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-7) Mark Zuckerberg posted an apology afterwards, starting with "Sorry for the disruption today."[\[5\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-5)

## Rogers goes silent, 2022

At 4:43 in the morning on July 8, 2022, Rogers, one of Canada's biggest phone and internet companies, was partway through an upgrade of its network. A small code change removed a filter that normally limited how much routing information flowed into the company's central equipment.[\[8\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-8)

Without the filter, that equipment was flooded with far more information than it could hold. It ran out of memory and shut down, and Rogers' whole network went with it.[\[8\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-8) A review done later for Canada's telecom regulator put the number of affected customers at more than 12 million.[\[9\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-9)

Phones stopped working. So did home internet. Rogers customers could not call 911. Interac, Canada's debit card system, also broke, which left shoppers and businesses struggling to pay and get paid.[\[8\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-8)

Finding the cause took a long time. Several changes had been made that day, and Rogers misread what had gone wrong for about 14 hours. The regulator's review said the crash could have been prevented if the central equipment had a basic limit on how much it would accept.[\[9\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-9) Overall, the outage lasted more than a day.[\[9\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-9)

## Amazon's Virginia problem, 2025

Shortly before midnight Pacific time on October 19, 2025, two automated systems inside Amazon's cloud, in the same northern Virginia region as in 2017, tripped over each other. Their job was to keep an address list up to date for one of Amazon's most widely used databases, a bit like a phone book. Because of a timing bug, the entry for that database ended up blank.[\[10\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-10)

Apps that needed the database could not find it. Then other Amazon services that relied on those apps began failing too, and the trouble spread outward for hours.[\[10\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-10)

Snapchat, Reddit, Venmo, Robinhood, Disney+, Lyft and the McDonald's app all had problems. Ring doorbells stopped working. Fortnite and Roblox players were stuck. Delta customers could not check in, and in Britain, Lloyds Banking Group and the government's tax office were hit.[\[11\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-11) Amazon's own shopping site and warehouse systems had trouble as well.[\[11\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-11)

Amazon said the address problem itself was fixed within about three hours. Cleaning up everything it had knocked over took much longer, and Amazon put the end of the event at 2:20 p.m. the next afternoon, more than 14 hours after it began.[\[10\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-10)

## Cloudflare's oversized file, 2025

Less than a month later, on November 18, 2025, it was Cloudflare's turn. Visits to a huge number of websites pass through Cloudflare's network first, which protects the sites and speeds them up. At 11:05 in the morning UK time, someone changed who was allowed to see what inside one of its databases.[\[12\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-12)

That small change caused the database to repeat itself. A file Cloudflare uses to spot bad automated traffic doubled in size. The software that read the file was built to handle about 200 entries, and normally used around 60. The new file went over the limit, the software crashed, and visitors to sites everywhere saw an error page blaming Cloudflare.[\[12\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-12)

The team first assumed it was under a massive attack. It did not help that the file was rebuilt every five minutes, so the network kept recovering and failing again. It helped even less that Cloudflare's status page, which runs elsewhere, went down at the same moment by pure coincidence.[\[12\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-12)

Most traffic was flowing again within about three hours, and everything was back to normal after nearly six. Matthew Prince, who wrote Cloudflare's account, called it the company's worst outage since 2019 and wrote, "I would like to apologize for the pain we caused the Internet today."[\[12\]](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#source-12)

## Sources

1. Fortune, David Meyer — “[From WhiteHouse.gov to Reddit: Fastly CDN outage knocks out sites and services around the world](https://fortune.com/2021/06/08/fastly-outage-cdn-services-internet/)” (June 2021). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-1-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-1-2)[↩3](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-1-3)[↩4](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-1-4)
2. Amazon Web Services — “[Summary of the Amazon S3 Service Disruption in the Northern Virginia (US-EAST-1) Region](https://aws.amazon.com/message/41926/)” (March 2017). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-2-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-2-2)[↩3](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-2-3)
3. The Register, Shaun Nichols — “[AWS's S3 outage was so bad Amazon couldn't get into its own dashboard to warn the world](https://www.theregister.com/2017/03/01/aws_s3_outage/)” (March 2017). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-3-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-3-2)
4. Fastly, Nick Rockwell — “[Summary of June 8 outage](https://www.fastly.com/blog/summary-of-june-8-outage)” (June 2021). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-4-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-4-2)[↩3](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-4-3)[↩4](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-4-4)[↩5](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-4-5)
5. CNBC, Sam Shead — “[Facebook says sorry for mass outage and reveals why it happened](https://www.cnbc.com/2021/10/05/facebook-says-sorry-for-mass-outage-and-reveals-why-it-happened.html)” (October 2021). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-5-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-5-2)
6. Engineering at Meta, Santosh Janardhan — “[More details about the October 4 outage](https://engineering.fb.com/2021/10/05/networking-traffic/outage-details/)” (October 2021). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-6-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-6-2)
7. NPR (The Associated Press) — “[Facebook, WhatsApp, Instagram suffer worldwide outage](https://www.npr.org/2021/10/04/1043098635/facebook-whatsapp-instagram-outage)” (October 2021). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-7-1)
8. The Globe and Mail, Alexandra Posadzki — “[How a coding error caused Rogers outage that left millions without service](https://www.theglobeandmail.com/business/article-how-a-coding-error-caused-rogers-outage-that-left-millions-without/)” (July 2022). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-8-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-8-2)[↩3](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-8-3)
9. BNN Bloomberg (The Canadian Press) — “[CRTC posts independent report on Rogers outage, says company made necessary changes](https://www.bnnbloomberg.ca/business/company-news/2024/07/05/crtc-posts-independent-report-on-rogers-outage-says-company-made-necessary-changes/)” (July 2024). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-9-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-9-2)[↩3](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-9-3)
10. Amazon Web Services — “[Summary of the Amazon DynamoDB Service Disruption in the Northern Virginia (US-EAST-1) Region](https://aws.amazon.com/message/101925/)” (October 2025). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-10-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-10-2)[↩3](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-10-3)
11. CNBC, Annie Palmer — “[AWS services recover after daylong outage hits major sites](https://www.cnbc.com/2025/10/20/amazon-web-services-outage-takes-down-major-websites.html)” (October 2025). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-11-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-11-2)
12. Cloudflare, Matthew Prince — “[Cloudflare outage on November 18, 2025](https://blog.cloudflare.com/18-november-2025-outage/)” (November 2025). [↩](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-12-1)[↩2](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-12-2)[↩3](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-12-3)[↩4](https://cloudchangelog.com/articles/small-mistakes-big-internet-outages#cite-12-4)

## More articles

- [Pentium FDIV Bug: The Chip That Got Division Wrong](https://cloudchangelog.com/articles/pentium-fdiv-bug)

  A math professor found that Intel's Pentium chip got some division wrong, and Intel's first answer made things worse.
- [Mars Climate Orbiter: How a Unit Mix-Up Lost a Spacecraft](https://cloudchangelog.com/articles/mars-climate-orbiter-unit-mix-up)

  A NASA spacecraft flew for nine months, then was lost within minutes of reaching Mars, because one file used pounds where everyone expected metric units.
- [A Short History of the Changelog](https://cloudchangelog.com/articles/history-of-the-changelog)

  How a plain list of software changes went from a programmer's private diary to app store jokes and pages that investors follow.

Get a free daily, weekly or monthly summary by email.

## Contact

- [YouTube](https://www.youtube.com/@cloudchangelog)

```json
{
  "@context": "https://schema.org",
  "@type": "Article",
  "author": {
    "@type": "Organization",
    "name": "Cloud Changelog"
  },
  "citation": [
    {
      "@type": "CreativeWork",
      "name": "From WhiteHouse.gov to Reddit: Fastly CDN outage knocks out sites and services around the world",
      "publisher": {
        "@type": "Organization",
        "name": "Fortune, David Meyer"
      },
      "url": "https://fortune.com/2021/06/08/fastly-outage-cdn-services-internet/"
    },
    {
      "@type": "CreativeWork",
      "name": "Summary of the Amazon S3 Service Disruption in the Northern Virginia (US-EAST-1) Region",
      "publisher": {
        "@type": "Organization",
        "name": "Amazon Web Services"
      },
      "url": "https://aws.amazon.com/message/41926/"
    },
    {
      "@type": "CreativeWork",
      "name": "AWS's S3 outage was so bad Amazon couldn't get into its own dashboard to warn the world",
      "publisher": {
        "@type": "Organization",
        "name": "The Register, Shaun Nichols"
      },
      "url": "https://www.theregister.com/2017/03/01/aws_s3_outage/"
    },
    {
      "@type": "CreativeWork",
      "name": "Summary of June 8 outage",
      "publisher": {
        "@type": "Organization",
        "name": "Fastly, Nick Rockwell"
      },
      "url": "https://www.fastly.com/blog/summary-of-june-8-outage"
    },
    {
      "@type": "CreativeWork",
      "name": "Facebook says sorry for mass outage and reveals why it happened",
      "publisher": {
        "@type": "Organization",
        "name": "CNBC, Sam Shead"
      },
      "url": "https://www.cnbc.com/2021/10/05/facebook-says-sorry-for-mass-outage-and-reveals-why-it-happened.html"
    },
    {
      "@type": "CreativeWork",
      "name": "More details about the October 4 outage",
      "publisher": {
        "@type": "Organization",
        "name": "Engineering at Meta, Santosh Janardhan"
      },
      "url": "https://engineering.fb.com/2021/10/05/networking-traffic/outage-details/"
    },
    {
      "@type": "CreativeWork",
      "name": "Facebook, WhatsApp, Instagram suffer worldwide outage",
      "publisher": {
        "@type": "Organization",
        "name": "NPR (The Associated Press)"
      },
      "url": "https://www.npr.org/2021/10/04/1043098635/facebook-whatsapp-instagram-outage"
    },
    {
      "@type": "CreativeWork",
      "name": "How a coding error caused Rogers outage that left millions without service",
      "publisher": {
        "@type": "Organization",
        "name": "The Globe and Mail, Alexandra Posadzki"
      },
      "url": "https://www.theglobeandmail.com/business/article-how-a-coding-error-caused-rogers-outage-that-left-millions-without/"
    },
    {
      "@type": "CreativeWork",
      "name": "CRTC posts independent report on Rogers outage, says company made necessary changes",
      "publisher": {
        "@type": "Organization",
        "name": "BNN Bloomberg (The Canadian Press)"
      },
      "url": "https://www.bnnbloomberg.ca/business/company-news/2024/07/05/crtc-posts-independent-report-on-rogers-outage-says-company-made-necessary-changes/"
    },
    {
      "@type": "CreativeWork",
      "name": "Summary of the Amazon DynamoDB Service Disruption in the Northern Virginia (US-EAST-1) Region",
      "publisher": {
        "@type": "Organization",
        "name": "Amazon Web Services"
      },
      "url": "https://aws.amazon.com/message/101925/"
    },
    {
      "@type": "CreativeWork",
      "name": "AWS services recover after daylong outage hits major sites",
      "publisher": {
        "@type": "Organization",
        "name": "CNBC, Annie Palmer"
      },
      "url": "https://www.cnbc.com/2025/10/20/amazon-web-services-outage-takes-down-major-websites.html"
    },
    {
      "@type": "CreativeWork",
      "name": "Cloudflare outage on November 18, 2025",
      "publisher": {
        "@type": "Organization",
        "name": "Cloudflare, Matthew Prince"
      },
      "url": "https://blog.cloudflare.com/18-november-2025-outage/"
    }
  ],
  "dateModified": "2026-11-19T14:05:00.000Z",
  "datePublished": "2026-11-19T14:05:00.000Z",
  "description": "A typo at Amazon, a setting at Fastly, a cleanup at Rogers. Six true stories of small slips at single companies that knocked out services for millions.",
  "headline": "Small Mistakes That Took Down Big Parts of the Internet",
  "image": "https://cloudchangelog.com/_files/3LKjxEzy7iiSV3FYehOlOemjcXv/ixYeFTqn5XtfXnXROfwGkQ/original.png",
  "mainEntityOfPage": "https://cloudchangelog.com/articles/small-mistakes-big-internet-outages",
  "publisher": {
    "@type": "Organization",
    "name": "Cloud Changelog"
  }
}
```
