“Last night the internet suddenly stopped working…” “X (Twitter), ChatGPT, even Discord — none of them would load!”

On the night of November 18, 2025, many people experienced exactly this. It wasn’t your phone or Wi-Fi — it was a global-scale Cloudflare outage that disrupted countless services.

Here is the timeline, the surprising cause, and what we can learn from it.

1. Timeline of the “3 Hours of Silence”

Incident Time and Impact

• Date: November 18, 2025 (Tue), from around 20:20 JST
• Symptoms: “502 Bad Gateway” errors, apps failing to load
• Major services affected:
  • X (Twitter)
  • ChatGPT (OpenAI)
  • Discord
  • Canva
  • Spotify
  • Millions of other websites globally

DownDetector charts spiked vertically — it was an unprecedented event.

Why Did Everything Go Down at Once?

All these services relied on Cloudflare’s security infrastructure. Cloudflare is like a massive traffic intersection — when all the traffic lights turned red simultaneously, every vehicle on every road ground to a halt.

2. The Cause: Not a Cyber Attack

Surprisingly, the cause was an internal system bug — not a cyber attack.

The Culprit: A Bot Protection Configuration File

Cloudflare’s “Bot Management” feature uses configuration files to block malicious bots. During an update, the feature file grew unexpectedly large due to unforeseen system behavior. When Cloudflare’s global servers tried to load this enormous file, they crashed.

Ironically, an update meant to enhance security ended up taking down the system.

3. Lessons and Countermeasures

What Users Can Do

• Recognize it is not just you: Check Cloudflare Status or news sites. Knowing it is a global issue saves you from pointless router restarts.
• Wait for recovery: Infrastructure-level outages can only be fixed by engineers on the other end.

What Site Operators Can Do

• Don’t make unnecessary changes: Changing server settings or DNS during an outage can cause secondary issues after recovery.
• Have emergency communication channels: Maintain multiple ways (SNS, email newsletters) to inform users if your site goes down — especially since X might also be down.

Summary

This outage highlighted the fragility of an internet where so many services depend on a single company’s infrastructure. But it also demonstrated the rapid response of Cloudflare’s engineering team.

• Cause: A bug from an oversized bot protection configuration file
• Not a cyber attack
• Fully recovered

For those interested in more technical details, Cloudflare’s official blog post is available.