5 hours ago
What is the recommended supported method for a consistent whole-volume backup of a nearly full Railway volume containing SQLite/WAL and related files? Does the documented 50% backup capacity limit prevent backups above 50% used, and is there an alternative without resizing or deleting data? How can all writers, including background workers, be quiesced, or a stopped service volume exported without starting the app? Can a complete backup be downloaded with checksums/manifests and restored separately for independent verification without overwriting the source? What consistency guarantees, downtime, and storage/export/egress/restore charges apply? Please distinguish free and billable options. This is a generic workflow question, not a request to change a service. No service identifiers, credentials, logs, or application data are included. Thank you.
1 Reply
5 hours ago
Native volume backups are the supported method. Each one is a point-in-time snapshot of the whole volume, so a SQLite database and its WAL files are captured together. This is crash consistency, not application consistency: we do not pause your app or flush its memory. For an application-consistent copy, you need to stop or quiesce every writer yourself first.
Manual backups are limited to 50% of the volume's total size, and the documented way past that is growing the volume. Scheduled backups (daily, weekly, monthly) also exist, but we have not verified whether the same limit applies to them.
A restore creates a new volume in the same project and environment and redeploys the service. The previous volume stays in the project, unmounted, rather than being overwritten. Because it redeploys, a restore is not a read-only test. Native backups cannot be downloaded as an archive and do not include checksums or manifests.
To export files yourself, use railway volume browse/railway volume files, or sftp/scp via railway ssh. These all go through a service that has the volume attached, so we have not confirmed a way to export while that service is stopped. Checksums would come from your own tooling.
Redeploying a service with a volume involves brief downtime. Backups are billed at the volume storage rate for their incremental data only (pricing).
Status changed to Awaiting User Response Railway • about 5 hours ago