Google Cloud Accidentally Suspends Railway for 10 Hours

追加: 2026-05-24

[00:00:00]On May 19th, around 10:10 p.m. UTC, Google Cloud suspended Railway's account in an automated action that swept up many other customers at the same time.

[00:00:10]There was no warning.

[00:00:12]Railway is a developer platform that hosts apps and databases for thousands of companies. They lost their dashboard, their API, and even their AWS bare metal workloads because those workloads depended on cached network routes from Google Cloud.

[00:00:29]Google restored access within just 20 minutes, but the damage was already done.

[00:00:34]Cached routes expired anyway, and full recovery took nearly 10 hours.

[00:00:40]Railway's incident report says the suspension was applied incorrectly.

[00:00:45]Google has not publicly explained what triggered it, and press press requests for comment have gotten nothing useful so far.

[00:00:53]Railway's report takes full responsibility for the architecture that let one provider cascade into a platform-wide outage, and lays out a plan to remove the single cloud dependency.

[00:01:06]"Our customers don't care if it is Google," a a Railway engineer told The Register. "We have to own our own uptime."

[00:01:14]When you depend on one cloud, your uptime depends on that cloud entirely.