Skip to content
ramira

13 November 2026 · 6 min · Web and mobile apps

The global Microsoft outage of 19 July 2024: timeline, cause, lessons

On Friday 19 July 2024, millions of Windows computers showed a blue error screen at the same time, all over the world. Windows itself was not the cause. The cause was an update to Falcon, CrowdStrike’s security software for businesses. Microsoft counted 8.5 million affected devices. Here is what happened, hour by hour, and what a small business can learn from it.

The baggage claim hall at New York’s LaGuardia airport: every display screen shows a blue error screen.
LaGuardia Airport (New York), on the morning of 19 July 2024. Photo: Smishra1, CC BY-SA 4.0, via Wikimedia Commons.

What happened, in one sentence

CrowdStrike sent a faulty configuration file to its Windows customers, and that file crashed the computer as soon as it started. The computer restarted, read the file again and crashed again. It looped on the blue screen. So it could not be fixed remotely.

The timeline

Times are in UTC, as in CrowdStrike’s reports. Add one hour for London and two hours for Paris.

Date and time (UTC) What happens
February 2024 CrowdStrike releases version 7.11 of its sensor. It adds a new type of detection rule. This rule type defines 21 fields.
19 July 2024, 04:09 CrowdStrike pushes an update of “Channel File 291” to Windows computers.
19 July 2024, from 04:09 Computers that receive the file crash and show the blue screen. Airports, banks, hospitals and TV channels are hit.
19 July 2024, 05:27 CrowdStrike fixes the file. Computers that have already crashed stay stuck.
19 July 2024, during the day CrowdStrike states that it is not a cyberattack. It publishes the manual repair steps.
20 July 2024 Microsoft estimates that 8.5 million Windows devices are affected. It publishes a USB recovery tool.
25 July 2024 More than 97% of Windows computers protected by CrowdStrike are back online, according to its CEO George Kurtz.
6 August 2024 CrowdStrike publishes its full root cause analysis.

The real cause

CrowdStrike’s software runs at the heart of Windows, at the same level as the system. An error there does not crash one application. It crashes the whole computer.

The 6 August 2024 report explains the bug. The new rule type expected 21 values. The code that called it supplied only 20. On 19 July, a rule read the 21st value for the first time. It read memory that did not exist. Windows stopped dead.

CrowdStrike’s checking tool missed the error. And the update went to all customers at once, without being tried on a small group first.

Only Windows computers running Falcon (version 7.11 or later) that were online between 04:09 and 05:27 UTC crashed. Macs and Linux servers were not affected.

Who was hit

Microsoft estimated that 8.5 million devices were affected, less than 1% of all Windows machines. That is small as a percentage. But these computers belonged to large companies, the ones that buy CrowdStrike. And they often sat in key places.

  • Air travel. 5,078 flights were cancelled worldwide that day, 4.6% of scheduled flights. Delta Air Lines cancelled more than 7,000 flights over several days. It put the cost at about 500 million dollars.
  • France. At Paris-Charles de Gaulle and Orly, some airlines had check-in problems. The Paris 2024 Olympics, which started a week later, saw accreditation handouts disrupted. Channels of the Canal+ group were also hit.
  • Health and emergency services. US hospitals cancelled operations. Several 911 emergency lines went down in the United States.
  • Banks and retail. Banks had services cut. Self-checkouts and payment terminals stopped.
A self-checkout screen in a French supermarket shows the blue error screen; a sheet of paper taped over it says the scanners are out of order.
A self-checkout in a French supermarket, 19 July 2024. Photo: Koupip, public domain (CC0), via Wikimedia Commons, cropped.

Why the repair took days

The fix came out 78 minutes after the faulty file. But a crashed computer could not receive it. Someone had to deal with each machine.

The method was always the same. Boot into safe mode or the Windows recovery environment. Delete the file whose name starts with “C-00000291”. Restart.

Two things slowed everything down:

  1. Someone had to be in front of the computer. For a company with thousands of machines across several sites, that takes time.
  2. The BitLocker recovery key was often needed. BitLocker encrypts the disk. Without its recovery key, you cannot touch the files. So you need it at hand for every machine.

On 20 July, Microsoft published a tool that builds a recovery USB stick. Microsoft also sent hundreds of engineers to help its customers. It worked with Google Cloud and Amazon Web Services on the best ways to recover.

In a hall at Washington Dulles airport, a large screen hanging above the gate signs shows the Windows blue error screen.
Washington Dulles Airport, 19 July 2024. Photo: reivax, CC BY-SA 2.0, via Wikimedia Commons.

What CrowdStrike changed

In its 6 August 2024 report, CrowdStrike announced several changes:

  • the number of fields is now checked before each release;
  • the software checks that it never reads outside the data it receives;
  • updates roll out in stages, first to a small group, then to everyone;
  • customers choose when and where updates are installed;
  • two outside security firms review the code.

That is the main lesson: an update that goes everywhere at once can break everything at once.

What a small business can do to suffer less

A small business cannot prevent an outage at a giant. It can shorten the time it stays stuck. Here is what we advise our clients.

  1. Write a one-page outage procedure. Who to call, which computer to fix first, how to warn customers. Print it. A procedure stored on the crashed computer is useless.
  2. Keep recovery keys off the computers. Store each machine’s BitLocker key in a password vault you can open from a phone, or on paper in a safe.
  3. Keep a way to work without your main computer. A phone or tablet that opens your online tools. A paper quote and invoice pad for a day. A second payment terminal from another provider, if taking payments is vital.
  4. Do not update every computer on the same day. For the updates you control, do one machine first, wait a day, then do the rest.
  5. Back up outside your computers, and test restores. A copy in the same office goes down with everything else. Read our method to back up Odoo automatically: it works for any software.
  6. Control who can change what. Fewer admin accounts means fewer ways for one mistake to spread. See how to set access rights in Odoo.
  7. Know where your data lives. An online tool depends on its host and its host’s suppliers. Ask before you sign.

In short

On 19 July 2024, a single configuration file knocked out 8.5 million Windows computers in 78 minutes. The repair took days, because every machine had to be handled. A small business cannot prevent this kind of outage. It can prepare with a paper procedure, well-stored recovery keys and a plan B to keep working.

Want tools that keep running when a supplier goes down? We build web and mobile apps designed for that, and we publish our own apps. For Odoo users, see our custom Odoo modules or contact us.

FAQ

Was it a Microsoft outage?

Not directly. The faulty update came from CrowdStrike, a security software vendor. But it crashed Windows, which is why it was widely called a global Microsoft outage. Microsoft helped its customers recover.

Were Macs and Linux computers affected?

No. CrowdStrike says Linux and macOS systems do not use the file involved. Only Windows computers running Falcon that were online between 04:09 and 05:27 UTC crashed.

Was a small business without CrowdStrike affected?

Its own computers did not crash. But it could still suffer from its suppliers’ outages: a bank, an airline, a payment terminal or an online tool hosted by a CrowdStrike customer.

How long did the outage last?

The faulty file was live for 78 minutes. The repair took days. By 25 July 2024, more than 97% of Windows computers protected by CrowdStrike were back online, according to its CEO.

Sources

Mohamed Rahmouni — founder of Ramira, app developer and Odoo consultant in Niort, France. More

Read next

An app, website or Odoo module project?

Describe it in a few lines: reply within 2 business days, with a first estimate.