ChatGPT, X, Spotify, Marmiton… A massive outage makes many sites and services on the Internet inaccessible

On Tuesday, November 18, the IT services giant Cloudflare suffered an unexpected outage which has made many websites inaccessible around the world. This large-scale bug hit major platforms at midday. Popular sites like Marmiton, Doctissimo, and Les Numériques have become inaccessible, as have crucial services such as ChatGPT or Downdetector; ironic, right?

The list of affected services is broad and also includes Facebook, cloud provider AWS (Amazon Web Services) — which itself suffered a breakdown in October — as well as entertainment platforms like Spotify and the video game League of Legends.

The impact of this incident is explained by the critical role of Cloudflare. The company is a central player in web infrastructure, providing essential services for millions of sites, including optimizing their performance and protecting against cyberattacks. The operator emphasizes that he answers “meets the Internet needs of millions of websites and processes an average of 81 million HTTP requests per second“.

Faced with these major problems on Cloudflare’s servers, a workaround solution quickly emerged for users: access via smartphone. Many affected online services — such as Instagram, ChatGPT, X, and Canva — remained accessible through their official mobile apps. The dysfunction was therefore mainly concentrated on access via the web. To use a service like ChatGPT whose site would be affected, it was therefore advisable to use the application or, if possible, to access it without being connected.

Recovery timeline

Cloudflare quickly communicated on the evolution of the situation:

At 12:48 p.m., the company announced that it was “aware of, and investigating, an issue affecting multiple customers,” without specifying the immediate cause.

As of 1:21 p.m., Cloudflare said it was “seeing a recovery in services,” while warning that customers may continue to see higher than normal error rates while remediation efforts continue.

At 2:30 p.m.: Cloudflare says “We made changes that got Cloudflare Access and WARP working again. Error levels for Access and WARP users returned to pre-incident rates“. And added: “we continue to work on restoring other services“.

At 3:35 p.m.: traffic on the majority of sites seems to have returned to normal.

We continue to monitor traffic recovery in real time.