On May 20th, Beijing time, the cloud deployment platform Railway experienced a major service outage. The root cause was Google Cloud blocking Railway’s accounts, rendering its dashboard, API, and internal network control plane inaccessible. All user services running on Google Cloud infrastructure also went down; reported errors included “no healthy upstream,” “unconditional drop overload,” and login failures. Workloads hosted on Railway Metal (its own physical servers) remained unaffected. The incident began around 6:29 a.m. Beijing time; Railway promptly reached out to Google Cloud support, but due to network issues on Google Cloud’s end, services couldn’t restart even after computing resources were restored. The recovery process took over seven hours in total.
During recovery, Railway limited build tasks for non-enterprise users to maintain infrastructure stability; enterprise deployments weren’t impacted. Around 2:14 p.m. Beijing time, Railway announced full service restoration and automatically triggered redeployments for workloads flagged as unhealthy. For any services still unresponsive, users could manually initiate redeployments via the dashboard or CLI. Railway stated it would publish a detailed postmortem report once stability was confirmed.