2 hours ago
What is the supported way to verify recovery of PostgreSQL data and an encrypted application-session directory on Railway without restarting production or replacing its mounted volume?
The verification must run with separate CPU and memory, must not connect to the application's external messaging provider, and must not export decryption keys outside their existing trust boundary. Is a read-only snapshot mount or another isolated workflow available?
How can spending for the verification alone be capped without risking shutdown of production workloads? Please clarify the required plan, any deployment or downtime involved, and which costs would apply before any changes are made.
I am asking for architecture guidance only, not authorizing an upgrade, resource creation, restore, or account change.
1 Reply
2 hours ago
We don't offer a read-only snapshot mount, and we have no supported isolated workflow for checking a volume backup while production stays unchanged. A volume backup restore always creates a new volume and stages a change that swaps it onto the original service. Deploying that change redeploys the production service. If you discard the staged change instead, production stays on its current volume, but the restored copy just sits unattached. We don't support attaching that copy to a separate service, browsing it, or exporting from it. Restores also only work inside the same project and environment.
Volume backups are point-in-time snapshots of the whole volume. They're crash-consistent, not application-consistent, so whether the data recovers cleanly depends on the database's own crash recovery. That's true for both PostgreSQL and the session directory if they share a volume.
You can't cap spending for one verification task by itself. Usage limits apply to the whole workspace. When compute usage reaches the hard limit, every workload in the workspace goes offline, production included. Custom email alerts only notify you and never stop anything. Per-service replica limits can cap a service's CPU and memory, but a service that goes over them can crash.
Status changed to Awaiting User Response Railway • about 2 hours ago