Internet paralyzed this morning: Cloudflare crashes again and takes millions of sites with it

Clearly, Cloudflare is accumulating failures. This December 5, around 10 a.m., the Californian company experienced its third major outage since the start of the year which seems to persist.

Platforms as diverse as LinkedIn, Perplexity, DeepL, Claude and Doctolib found themselves out of service, displaying 500 error messages. The irony reaches its height when we discover that Downdetector, the site supposed to record outages, itself collapsed for a moment, a victim of its own host.

Maintenance that goes wrong, again

The incident appears directly linked to a planned maintenance operation that went wrong. On its status page, Cloudflare confirms that it is investigating issues affecting its dashboard and APIs. “Cloudflare is currently investigating issues with the Cloudflare Dashboard and related APIs. Customers using the Cloudflare Dashboard/APIs are affected as requests may fail and/or errors may be displayed”soberly indicates the company.

The scenario furiously recalls the outage of November 18, itself caused by an internal configuration error. At the time, a bot management file had unexpectedly doubled in size, causing systems to collapse in a cascade. Matthew Prince, co-founder and CEO of Cloudflare, had to apologize.

This new failure still poses this embarrassing question: were the fixes applied after November really up to the task? For a service provider that manages between 15 and 20% of global web traffic, the repetition of similar incidents in the space of a few weeks reveals a very worrying fragility.