Google Cloud blocked Railway’s account on May 19, taking down its dashboard, API, edge network, and all customer workloads with no ETA for restoration.
Root cause: Google Cloud account block, not a Railway infrastructure fault internally – access was restored to upstream but workloads remained offline.
Railway’s dashboard, API, and internal network control plane all depend on the same GCP infrastructure with no apparent fallback.
As of May 20 01:23 UTC, the team was evaluating “alternative paths” to restore services while in direct contact with Google Cloud support.
Hacker News Comment Review
Commenters flagged a contradiction: Railway’s own blog claimed “you can’t build a cloud on another cloud” and touted running their own metal, yet the outage exposed full GCP dependency.
There is split opinion on fault assignment – some blame Google’s account-blocking pattern (also seen with a Korean government org), others argue Railway should have architected blast-radius isolation so one provider block cannot kill the entire control plane.
Trust damage was immediate and concrete: multiple commenters who were active or prospective Railway customers said the outage changed their evaluation, not just because of the outage itself but because of the total-service failure mode it revealed.
Notable Comments
@Avicebron: References a prior Railway incident involving backup keys stored in the prod database – pattern of single-point-of-failure decisions.
@miniman1337: Quotes Railway blog verbatim: “you can’t build a cloud on another cloud” – direct contradiction of today’s GCP-only failure.
