Internet Outage Atlas · Archive

THE
ARCHIVES

The full incident record, grouped by failure type and openable in place

All incidents in one place

This page holds the full atlas record. The homepage is selective. The timeline is for scanning by date. The archive is where every incident stays available in detail.

If you arrive from the timeline, the matching record opens automatically.

Filter the archive

Narrow the ledger by year, failure type, severity, or title. Source notes cite the research report and the narrative feature where that incident is covered there as well.

54 incidents shown
Internet Outage Atlas · Full Merged Research Report The Days the Internet Died
Routing failures

When bad maps become public reality

The service may still exist. The route to it is what disappears, leaks, or gets replaced by something false.

April 1997 BGP leak

AS7007 Route Leak

High Severity Open record

A misconfigured router at MAI Network Services originated a massive set of more-specific routes and polluted routing tables across the internet.

Why it spread

The event became an early proof that one broken routing announcement could destabilize far more than the network that sent it.

Failure type

Routing failures · BGP leak

Sources
  • Internet Outage Atlas · Full Merged Research Report
February 2008 BGP Route Hijack

Pakistan Telecom Blacks Out YouTube for the World

Critical Open record

Pakistan tried to block YouTube at home.

Why it spread

The route escaped, spread, and briefly blacked out YouTube for everyone else. A local censorship action became a global routing fact.

Failure type

Routing failures · BGP Route Hijack

Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
April 2014 Route hijack

Indosat BGP Hijack

High Severity Open record

An Indonesian provider briefly announced routes for large portions of the internet, diverting traffic that had nothing to do with it.

Why it spread

The incident showed how an operational mistake in one network can distort global reachability in minutes.

Failure type

Routing failures · Route hijack

Sources
  • Internet Outage Atlas · Full Merged Research Report
July 2015 Route leak

AxcelX and AWS Route Leak

High Severity Open record

Routes connected to AWS address space leaked outward and disrupted access to major sites and services.

Why it spread

The fault was not in application code. It was in the routing layer that decides where traffic goes at all.

Failure type

Routing failures · Route leak

Sources
  • Internet Outage Atlas · Full Merged Research Report
November 2018 Route leak

MainOne and Google Route Leak

High Severity Open record

A route leak involving MainOne and China Telecom redirected traffic for Google and other large services through unexpected paths.

Why it spread

It was a sharp demonstration of how brittle inter-network trust still is.

Failure type

Routing failures · Route leak

Sources
  • Internet Outage Atlas · Full Merged Research Report
June 2019 BGP Route Leak

Verizon Route Leak Disrupts 15 Percent of Global Internet Traffic

Critical Open record

A small provider leaked routes it did not own.

Why it spread

Verizon accepted them and propagated them. Cloudflare, Facebook, Google, and much more of the network got pulled into the mistake.

Failure type

Routing failures · BGP Route Leak

Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
October 2021 BGP / Platform

Facebook Global Blackout

Critical Open record

One backbone maintenance change withdrew Meta's BGP routes.

Why it spread

Facebook, Instagram, WhatsApp, and Messenger vanished at the same time. The harder part came next. The same failure also cut engineers off from some of the internal systems needed to fix it, which turned an outage into a recovery trap.

Failure type

Routing failures · BGP / Platform

Duration: approximately 6 hours · October 4, 2021
Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
2025 Route leak

Telekom Malaysia Route Leak

High Severity Open record

A modern route leak involving Telekom Malaysia preserved the same old lesson in contemporary form: routing mistakes still escape local intent and become international reachability problems very quickly.

Failure type

Routing failures · Route leak

Sources
  • Internet Outage Atlas · Full Merged Research Report
2026 BGP / address announcement failure

Cloudflare BYOIP BGP Outage

Critical Open record

A large-scale Cloudflare routing incident tied to BYOIP handling showed how reachability can still disappear at internet scale when address announcement logic goes wrong at a major provider edge.

Failure type

Routing failures · BGP / address announcement failure

Sources
  • Internet Outage Atlas · Full Merged Research Report
Naming, identity, and trust

When the service still exists but users cannot reach or trust it

Some failures do not knock servers offline. They break naming, authentication, or certificate trust, which is enough to make healthy systems feel dead.

October 2016 DNS / DDoS

Dyn DNS DDoS Attack

Critical Open record

Mirai used hacked cameras, routers, DVRs, and other junk devices to hammer Dyn's DNS infrastructure.

Why it spread

Twitter, Spotify, GitHub, Reddit, and much of the East Coast web started failing together. The point was not only the attack. It was how much of the visible web depended on one naming layer.

Failure type

Naming, identity, and trust · DNS / DDoS

Duration: approximately 8 hours · October 21, 2016
Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
2020 Cloud DNS

Azure DNS Outage

High Severity Open record

A separate Azure DNS incident in 2020 reinforced that naming failures recur even inside large cloud platforms, and that those failures can outrank the health of the underlying services they point to.

Failure type

Naming, identity, and trust · Cloud DNS

Sources
  • Internet Outage Atlas · Full Merged Research Report
May 2020 PKI expiration

Sectigo AddTrust Root Expiration

High Severity Open record

The expiration of the AddTrust root certificate triggered trust failures on legacy systems and broke connections that still depended on that chain.

Why it spread

A quiet certificate deadline turned into a visible service problem for older clients.

Failure type

Naming, identity, and trust · PKI expiration

Sources
  • Internet Outage Atlas · Full Merged Research Report
December 2020 Authentication Failure

Google Auth Outage Renders All Google Services Inaccessible

Critical Open record

A quota enforcement mistake during an auth migration knocked out Gmail, YouTube, Drive, and everything else tied to the same gate.

Why it spread

The apps were not the first problem. Access was.

Failure type

Naming, identity, and trust · Authentication Failure

Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
2021 Certificate expiration

Google Voice Expired TLS Certificate

High Severity Open record

An expired TLS certificate broke access to Google Voice and showed, again, that trust-chain maintenance is part of availability engineering rather than a side concern reserved for security teams.

Failure type

Naming, identity, and trust · Certificate expiration

Sources
  • Internet Outage Atlas · Full Merged Research Report
July 2021 Cloud DNS

Azure DNS Outage

High Severity Open record

A DNS-layer problem inside Azure interrupted name resolution and widened into a broader cloud-service disruption.

Why it spread

Parts of the visibility and status path became unreliable too, which made diagnosis harder for customers already in the dark.

Failure type

Naming, identity, and trust · Cloud DNS

Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
September 2021 Identity failure

Azure AD Key-Rotation Outage

Critical Open record

A signing-key problem became a long authentication outage across Microsoft 365, Teams, Exchange Online, and related services.

Why it spread

Systems were still there, but access to them was blocked by the gate in front.

Failure type

Naming, identity, and trust · Identity failure

Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
September 2021 PKI expiration

Let's Encrypt DST Root CA X3 Expiration

High Severity Open record

The expiration of DST Root CA X3 caused compatibility failures on older Android devices and legacy clients that still anchored trust there.

Why it spread

Modern infrastructure stayed up while part of the user base lost the ability to connect cleanly.

Failure type

Naming, identity, and trust · PKI expiration

Sources
  • Internet Outage Atlas · Full Merged Research Report
2023 Local network control

UAF DHCP Server Outage

Medium Severity Open record

A DHCP outage at the University of Alaska Fairbanks remains useful because it shows a bounded local-control failure that stayed local, which makes it a clean contrast against the much wider shared-layer incidents elsewhere in the atlas.

Failure type

Naming, identity, and trust · Local network control

Sources
  • Internet Outage Atlas · Full Merged Research Report
October 2025 Cloud / DNS

AWS DynamoDB DNS Failure

Critical Open record

DNS resolution for DynamoDB failed in us-east-1.

Why it spread

Disney+, Delta, Reddit, Robinhood, Roblox, and many other services went dark. The data was still there. The names stopped resolving. A naming failure overruled the resilience of the underlying system.

Failure type

Naming, identity, and trust · Cloud / DNS

Duration: approximately 15 hours · October 20, 2025
Sources
  • Internet Outage Atlas · Full Merged Research Report
Shared platforms and front doors

When unrelated services fail together

These incidents mattered because one shared provider or shared operational layer sat in front of many different products at once.

May 2021 SaaS DNS

Salesforce Multi-hour Outage

High Severity Open record

A DNS-related failure affected Salesforce services and disrupted the large body of business workflows built on top of them.

Why it spread

The outage hit as an enterprise dependency problem, not just an app problem.

Failure type

Shared platforms and front doors · SaaS DNS

Sources
  • Internet Outage Atlas · Full Merged Research Report
June 2021 CDN / Edge

Fastly Global Content Delivery Outage

Critical Open record

A dormant software bug sat in Fastly's network until a customer pushed a valid configuration change.

Why it spread

Within seconds, most of Fastly's global edge started returning errors. News sites, commerce platforms, and government pages disappeared together because the front door was more shared than it looked.

Failure type

Shared platforms and front doors · CDN / Edge

Duration: approximately 1 hour · June 8, 2021
Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
July 2021 DDoS mitigation

Akamai Prolexic Outage

High Severity Open record

A platform designed to preserve availability became the source of unavailability instead.

Why it spread

Customers depending on Prolexic lost service because the defensive layer itself failed under load.

Failure type

Shared platforms and front doors · DDoS mitigation

Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
July 2021 DNS / CDN

Akamai DNS Outage Silences FedEx, Airlines, and Major Banks

Critical Open record

A configuration update triggered a bug in Akamai Edge DNS and took down a long list of companies that looked unrelated until they failed at the same time.

Failure type

Shared platforms and front doors · DNS / CDN

Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
2025 Azure Front Door configuration

Microsoft 365 / Azure Global Outage

Critical Open record

A global Microsoft 365 and Azure outage tied to Azure Front Door configuration reinforced the atlas theme that the front door often fails harder than the applications behind it.

Failure type

Shared platforms and front doors · Azure Front Door configuration

Sources
  • Internet Outage Atlas · Full Merged Research Report
2025 Protective edge logic failure

Cloudflare Bot-management Outage

Critical Open record

An internal Cloudflare bot-management failure propagated widely because the protective layer itself sat in front of customer traffic at massive scale.

Why it spread

The case fits the broader pattern of defensive systems becoming shared failure systems.

Failure type

Shared platforms and front doors · Protective edge logic failure

Sources
  • Internet Outage Atlas · Full Merged Research Report
Cloud and control planes

When recovery tools and platform internals start failing too

The hardest cloud outages are not just service failures. They are outages where the systems needed to understand or recover the outage are also under stress.

April 2011 Cloud Infrastructure

AWS EC2 Failure Exposes Limits of Availability Zone Isolation

Critical Open record

A network upgrade misrouted EBS traffic and took volumes offline in us-east-1.

Why it spread

Reddit, Quora, Foursquare, and other services built too tightly around one zone lost their cushion fast.

Failure type

Cloud and control planes · Cloud Infrastructure

Sources
  • Internet Outage Atlas · Full Merged Research Report
November 2014 Cloud deploy failure

Azure Storage Outage

Critical Open record

Human error during a storage-system deployment led to a broad Azure outage and became one of the clearer early examples of control-plane mistakes causing large customer impact.

Why it spread

The trigger was routine. The spread was not.

Failure type

Cloud and control planes · Cloud deploy failure

Sources
  • Internet Outage Atlas · Full Merged Research Report
February 2017 Cloud / Human Error

AWS S3 US-East-1 Outage

Critical Open record

A mistyped debugging command removed more S3 capacity than intended.

Why it spread

Thousands of applications went down with it, including systems people did not realize depended on that region so heavily. The outage became a lasting example of how one control-plane mistake can turn a local action into a public outage.

Failure type

Cloud and control planes · Cloud / Human Error

Duration: approximately 4 hours · February 28, 2017
Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
June 2019 Cloud networking

Google Cloud Networking Outage

Critical Open record

A routine change cascaded through Google Cloud's networking systems and led to major traffic loss and degraded access across services.

Why it spread

The incident showed how internal reliability changes can widen into public unavailability.

Failure type

Cloud and control planes · Cloud networking

Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
December 2021 Cloud DNS and control plane

AWS US-East-1 Control-Plane Outage

Critical Open record

Internal networking and DNS issues in US-East-1 disrupted AWS services, Amazon devices, logistics systems, and third-party applications.

Why it spread

The region concentration problem was visible, but so was the depth of internal dependency inside the same region.

Failure type

Cloud and control planes · Cloud DNS and control plane

Sources
  • Internet Outage Atlas · Full Merged Research Report
2023 Regional cloud networking

GCP us-east4 Traffic Loss

High Severity Open record

Traffic loss in Google Cloud's us-east4 region highlighted how regional networking faults can still create large downstream application problems when many services quietly share the same cloud locality.

Failure type

Cloud and control planes · Regional cloud networking

Sources
  • Internet Outage Atlas · Full Merged Research Report
2026 Enterprise suite outage

Microsoft 365 Outage

Critical Open record

A long Microsoft 365 outage highlighted how deeply office coordination, messaging, documents, and identity have been consolidated into one operational dependency for many organizations.

Failure type

Cloud and control planes · Enterprise suite outage

Sources
  • Internet Outage Atlas · Full Merged Research Report
Platform and software cascades

When one release or one internal dependency spreads everywhere

These failures travel through shared software, internal coordination systems, or platform dependencies that turn one change into widespread operational loss.

December 2010 P2P platform failure

Skype Supernode Failure

High Severity Open record

A software problem destabilized Skype's supernode layer and broke service for a huge share of users, showing how coordination nodes inside distributed platforms can still become central failure points.

Failure type

Platform and software cascades · P2P platform failure

Sources
  • Internet Outage Atlas · Full Merged Research Report
August 2013 Platform failure

Google Global 5-minute Outage

High Severity Open record

A brief but iconic Google outage took major services offline at the same time and became a durable example of how concentrated platform ecosystems can vanish all at once, even during a short failure.

Failure type

Platform and software cascades · Platform failure

Sources
  • Internet Outage Atlas · Full Merged Research Report
October 2013 Commerce platform

Amazon.com Retail Outage

High Severity Open record

A high-profile Amazon retail outage showed how visible and immediate the impact becomes when a single commerce platform failure blocks browsing, purchasing, and order flow together.

Failure type

Platform and software cascades · Commerce platform

Sources
  • Internet Outage Atlas · Full Merged Research Report
June 2017 Software supply chain

NotPetya Global Outage

Critical Open record

NotPetya spread through a trusted software-update channel and crippled shipping, logistics, hospitals, and enterprise networks around the world.

Why it spread

It remains one of the clearest demonstrations of software supply chains acting like outage multipliers.

Failure type

Platform and software cascades · Software supply chain

Sources
  • Internet Outage Atlas · Full Merged Research Report
August 2020 Video platform

Zoom Partial Global Outage

High Severity Open record

Zoom experienced a broad service disruption during the period when remote work had made video infrastructure a daily dependency.

Why it spread

The incident showed how a platform that looks optional can become operationally central very quickly.

Failure type

Platform and software cascades · Video platform

Sources
  • Internet Outage Atlas · Full Merged Research Report
2020–2022 Storage exhaustion

Slack File-storage Outage

High Severity Open record

A file-storage failure inside Slack disrupted access to uploads and working materials, showing how collaboration platforms break not only when messaging fails but also when their attached operational data stops moving.

Failure type

Platform and software cascades · Storage exhaustion

Sources
  • Internet Outage Atlas · Full Merged Research Report
October 2021 Distributed systems failure

Roblox 73-hour Outage

Critical Open record

Roblox went down for roughly three days after failures involving internal service-discovery and data systems compounded across a highly interconnected platform.

Why it spread

The length of the outage made the recovery-path problem impossible to ignore.

Failure type

Platform and software cascades · Distributed systems failure

Sources
  • Internet Outage Atlas · Full Merged Research Report
February 2022 Collaboration platform

Slack Outage

High Severity Open record

Slack suffered a cascading failure involving database and cache systems, which disrupted messaging, connections, and workflow continuity for teams that depend on it as operating infrastructure.

Why it spread

Recovery was shaped by how many internal pieces were failing together.

Failure type

Platform and software cascades · Collaboration platform

Sources
  • Internet Outage Atlas · Full Merged Research Report
2024 Third-party roaming dependency

AT&T / T-Mobile / Verizon Roaming Outage

Critical Open record

A shared roaming dependency disrupted multiple major U.S.

Why it spread

carriers at once, making the outage notable less for any one brand than for the hidden third-party relationship that linked them together.

Failure type

Platform and software cascades · Third-party roaming dependency

Sources
  • Internet Outage Atlas · Full Merged Research Report
March 2024 Platform ecosystem outage

Meta (Facebook / Instagram) Outage

Critical Open record

A broad Meta outage affecting Facebook and Instagram showed that even without a long root-cause disclosure, the operational story remains the same: concentrated social platforms fail at the scale of their audience.

Failure type

Platform and software cascades · Platform ecosystem outage

Sources
  • Internet Outage Atlas · Full Merged Research Report
July 2024 Software Supply Chain

CrowdStrike Falcon Global Outage

Critical Open record

A routine CrowdStrike update shipped a bad configuration file and crashed Windows at kernel level on an estimated 8.5 million devices.

Why it spread

Airlines could not board passengers. Hospitals switched to paper. Banks shut down systems. Recovery was slow because every broken machine needed hands-on work.

Failure type

Platform and software cascades · Software Supply Chain

Duration: recovery spanning 10 days · July 19, 2024
Sources
  • Internet Outage Atlas · Full Merged Research Report
  • The Days the Internet Died
2025 Developer platform outage

GitHub Outage

High Severity Open record

A GitHub outage disrupted repository operations and development workflows at a layer many teams now treat as critical infrastructure rather than an optional collaboration tool.

Failure type

Platform and software cascades · Developer platform outage

Sources
  • Internet Outage Atlas · Full Merged Research Report
Physical and public infrastructure

When geography and public systems stop being abstract

The cloud still depends on buildings, cables, power, carrier paths, and public-safety infrastructure that can fail in the same event.

August 2005 Telecom infrastructure

Hurricane Katrina Telecom Failures

Critical Open record

Katrina destroyed fiber paths, towers, power, and fuel logistics together, collapsing multiple redundant systems at once.

Why it spread

It remains one of the clearest examples of geography overpowering abstract redundancy claims.

Failure type

Physical and public infrastructure · Telecom infrastructure

Sources
  • Internet Outage Atlas · Full Merged Research Report
February 2008 Datacenter power

The Planet Houston Outage

High Severity Open record

A power failure at The Planet's Houston datacenter exposed how fragile backup systems can be when they are tested under real pressure.

Why it spread

Thousands of hosted servers went dark and recovery stretched across days.

Failure type

Physical and public infrastructure · Datacenter power

Sources
  • Internet Outage Atlas · Full Merged Research Report
2014 Backbone fiber failure

Level 3 Fiber Outage

High Severity Open record

A backbone fiber disruption at Level 3 highlighted how physical transport failures can still cascade into broad connectivity problems across downstream networks that rely on the same paths.

Failure type

Physical and public infrastructure · Backbone fiber failure

Sources
  • Internet Outage Atlas · Full Merged Research Report
June 2018 Fiber cut

Comcast Fiber Cut Outage

High Severity Open record

A large Comcast outage traced back to physical infrastructure damage and showed how ordinary cable-path failures can still produce wide consumer and enterprise impact.

Why it spread

The cloud did not make the fiber less real.

Failure type

Physical and public infrastructure · Fiber cut

Sources
  • Internet Outage Atlas · Full Merged Research Report
July 2022 Telecom Core Failure

Rogers Canada: 12 Million Without Service, Including 911

Critical Open record

A core network upgrade removed a critical filter and sent traffic into the wrong place at the wrong scale.

Why it spread

Rogers collapsed under the load, taking mobile service, internet access, and 911 with it for millions of people.

Failure type

Physical and public infrastructure · Telecom Core Failure

Sources
  • Internet Outage Atlas · Full Merged Research Report
2024 Mobile carrier outage

Verizon Mobile Outage

Critical Open record

A major Verizon mobile outage underscored how quickly carrier failures still spill into daily public life once voice, data, authentication, and payment flows all assume cellular reachability.

Failure type

Physical and public infrastructure · Mobile carrier outage

Sources
  • Internet Outage Atlas · Full Merged Research Report
2026 Mobile carrier outage

Verizon Mobile Outage

High Severity Open record

A later Verizon mobile outage, even with thinner public disclosure, remains useful in the atlas because it reinforces how dependent daily communications and service access remain on a small number of carrier systems.

Failure type

Physical and public infrastructure · Mobile carrier outage

Sources
  • Internet Outage Atlas · Full Merged Research Report
2026 Physical attack on cloud infrastructure

AWS Middle East Drone-strike Outage

Critical Open record

This incident is preserved because it forces the cloud back into physical reality: regional availability ultimately depends on facilities, geography, power, and security on the ground.

Failure type

Physical and public infrastructure · Physical attack on cloud infrastructure

Sources
  • Internet Outage Atlas · Full Merged Research Report