Railway restores deploys after GCP outage

Railway said on May 20 that deployments were working again after a Google Cloud account suspension triggered an outage that spread beyond the company’s Google Cloud footprint and disrupted customer workloads across the platform. The company said users could redeploy to route code to healthy machines while it continued restoring affected Google Cloud hosts. Railway also published a public incident report, a Central Station FAQ and live status updates describing the failure and the recovery steps. ### How did a Google Cloud issue turn into a platform-wide Railway outage? Railway said Google Cloud placed its production account into a suspended or restricted status at about 22:20 UTC on May 19, taking offline the company’s API, control plane, databases and compute infrastructure hosted on Google Cloud. The company said that immediately produced 503 errors on the dashboard and API and blocked logins. (blog.railway.com) Railway said the outage then spread because its edge proxies depended on a Google Cloud-hosted control plane API to populate routing tables. As cached routes expired, workloads on Railway Metal and AWS burst capacity also became unreachable, and users began seeing 404 errors because the network control plane could no longer resolve routes to active instances. (blog.railway.com) ### Why were deploys blocked even after some infrastructure came back? Railway said builds and deployments were blocked across the platform while it restored individual services in Google Cloud and brought its infrastructure back online. The company said it then drained a large backlog of queued deploys gradually to avoid overloading the platform. (blog.railway.com) The status page on May 20 showed a separate degraded-performance notice for slow builds on Railway Metal, with free, trial and hobby builds temporarily paused to reduce queue backlog while Pro and higher plans continued to process. Railway said running services were not affected by that build delay notice. ### What did Railway tell customers to do right now? Angelo Saraceno, posting in Railway’s Central Station forum, said users could redeploy and Railway would route their code to a healthy machine if they were not yet fully recovered. (blog.railway.com) He also said the support team was working to restore workloads on Google Cloud hosts and warned that GitHub rate limits were slowing parts of the build pipeline recovery. (status.railway.com) Railway’s incident report said GitHub also began rate-limiting the company’s OAuth and webhook integrations during recovery, temporarily blocking some logins and builds as caches were cleared after the Google Cloud outage. ### What does Railway say it got wrong in its architecture? Railway said it had moved fully onto its own metal infrastructure in March 2025 but kept its API and database on Google Cloud because it considered that risk acceptable at the time. (station.railway.com) In the Central Station FAQ, the company said it did not expect an automated enforcement action to remove the cloud account. (blog.railway.com) In its May 20 incident report, Railway said it took “full responsibility” for the architectural decisions that let a single upstream provider action cascade into a platform-wide outage. The report said the company was detailing changes intended to prevent a repeat, including reducing dependence on a single provider in the critical path. (station.railway.com) ### Where is Railway publishing the follow-up? Railway said on May 20 that the incident report reflected what it knew at publication time and could be updated pending Google Cloud’s internal review. The company also said in Central Station that it would publish an official postmortem once recovery was complete. The next updates are being posted in three places Railway named publicly: its status page, its Central Station FAQ thread and its blog post titled “Incident Report: May 19, 2026 - GCP Account Suspension.” (status.railway.com) (blog.railway.com)