Cloudflare confirmed that the massive outage that affected many services yesterday had nothing to do with a security incident, and most importantly, no data was lost.
The problem, although serious, has now been almost completely resolved. It all began at 17:52 UTC, when one of the key components of its infrastructure (Workers KV, a key-value data storage system) went completely offline. This caused a chain reaction of failures in several of the computing and artificial intelligence services that rely on that technology.
In case you're not familiar with it, Workers KV is like a huge distributed database used within Cloudflare Workers, the platform that allows code to be executed without the need for servers. It's an essential piece, and when something fails there, it's felt in many other parts of the system.
Furthermore, this outage not only affected Cloudflare, but also other widely used services, including Google Cloud Platform, which explains why so many users noticed errors on different platforms.
KV error rate of workers during the incident (Source: Cloudflare)
Read more: Massive Internet Outage: Google, Cloudflare, Character.AI, Twitch...
Cloudflare explains its major outage and what it plans to do to prevent another one
After yesterday's major outage on Thursday, June 12, 2025, Cloudflare published a kind of “autopsy” of the incident, explaining exactly what happened, how long it lasted, and what they are going to do to prevent it from happening again.
How long did it last?
The outage lasted almost two and a half hours. It all started with a failure in the storage system used by Workers KV, an essential part of the engine that powers many of Cloudflare's services.
And what was the cause?
The fault lay partly with an external cloud provider that had its own problem. That third-party cloud is what backs up part of Workers KV's storage. When it failed, it brought down several services that depend on that structure to function.
Cloudflare explains it this way:
“The cause was a failure in the storage infrastructure we use for Workers KV. This infrastructure is critical, as it is used for configuration, authentication, and content delivery in many of our products.”
Which services were affected?
Basically, almost the entire Cloudflare ecosystem suffered in some way. Here is a clear summary:
- Workers KV: had a 90% error rate. If it wasn't in the cache, it simply didn't work.
- Access, WARP, and Gateway: serious failures in login, authentication, and device management. WARP couldn't register new devices.
- Control panel and CAPTCHA (Turnstile): failures in login and verification. An “emergency switch” had to be used, which introduced the risk of token reuse.
- Remote browser (Browser Isolation): sessions couldn't be started or maintained due to related errors.
- Stream, Images, and Pages: streams went down, image uploads failed completely, and Pages publications crashed with almost total errors.
- Workers AI and AutoRAG: were completely out of service.
- Durable objects, D1, and queues: high errors or completely inactive services.
- Realtime and AI Gateway: were also almost completely unusable.
- Zaraz and Workers Assets: failures when loading configurations, although the impact on the end user was limited.
- CDN and Workers Builds: high latency, regional errors, and total failures in new builds.
In short, it was a technical disaster of the kind that leaves its mark.
What will Cloudflare do now?
Cloudflare did not sit idly by. They have already announced a series of important changes to improve the resilience of their systems:
-
Less dependence on third parties: they will begin moving Workers KV to Cloudflare R2, their own object storage system. This gives them more control and less risk.
-
Layers of protection between services: they will implement barriers to prevent a failure in one area from affecting all others.
-
Better progressive recovery: they are developing tools to restore services in stages, avoiding those traffic spikes that crash systems in recovery.
Conclusion: A good scare, but also a good lesson
Although it was a significant outage and very annoying for millions of users and businesses, Cloudflare responded quickly, was transparent, and is already taking concrete steps to strengthen its infrastructure.
Incidents like this, while undesirable, serve to improve, and most importantly, there was no hacking or data loss. Just a broken dependency that highlighted how fragile even the largest Internet platforms can be.