Google Cloud blocked Railway’s account on May 19, causing a major outage affecting the dashboard, API, edge network, and all GCP-hosted workloads.
Key Takeaways
Symptoms included “no healthy upstream”, “unconditional drop overload”, login failures, and dashboard inaccessibility.
Railway confirmed GCP as the sole cause at 23:37 UTC; the block affected the dashboard, API, and internal network control plane.
GCP access was partially restored within ~15 minutes of identification, but full workload restoration had no ETA as of 00:37 UTC May 20.
Railway’s own domain (railway.com) went down, exposing a single-cloud dependency with no apparent failover for control-plane infrastructure.
Hacker News Comment Review
Strong consensus that GCP’s account-blocking behavior without prior contact is a known, recurring risk that makes it unsuitable for production infrastructure despite its technical quality.
Commenters split on whether the trigger was automated abuse detection (Railway IPs are reportedly a significant spam source) or GCP’s opaque support processes, with no official cause given.
GCP’s support structure drew heavy criticism: slow escalation, no proactive outreach for high-spend accounts, contrasted unfavorably with AWS TAM behavior.
Notable Comments
@BitWiseVibe: argues Railway has “horrible abuse prevention” and that spam volume from Railway IPs may have triggered the block.
@brokenodo: migrated to Render mid-incident with DNS cutover in 1 minute and full restore in ~30 minutes, a concrete data point on switching cost.
