Skip to content
Data residency — storage region for volume backups and PITR archive
Anonymous
OP

a month ago

We operate a Postgres service deployed in europe-west4 (Netherlands).

For a regulatory filing and a GDPR / Saudi PDPL data-residency assessment, we need written confirmation of the following:

  1. The region in which scheduled and on-demand volume backups for that service are physically stored.
  2. The region in which the point-in-time recovery archive for that service is physically stored.
  3. Whether either is replicated to, cached in, or processed in a United States region at any point.
  4. The underlying infrastructure provider and region for each.

Your documentation states that backups can only be restored into the same project and environment, but does not state where they are stored, and the compliance page does not publish a sub-processor or residency list.

A written answer we can retain for our records would be ideal; a link to a document stating it is equally good.

Solved

7 Replies

Railway
BOT

a month ago

Volume backups (scheduled and on-demand) are stored in the region the service deploys to, so a volume in EU West (Amsterdam, Netherlands) is backed up in EU West.

The Postgres point-in-time recovery archive (WAL segments and base backups) is also stored in the deploy region, in a Railway storage bucket. Bucket storage runs on Tigris, so for an EU West service that archive sits on Tigris in EU West.

Neither of those native features copies data to a US region. Two caveats belong in your assessment. First, all application, deploy, build, and HTTP logs are stored in US West regardless of deploy region, so anything your Postgres service writes to its logs leaves the EU. Second, we keep separate internal platform backups for disaster recovery, and the facts above are not a location inventory for those, so we can't state that no copy of any data ever exists outside the EU. We also don't offer a contractual EU-only processing commitment on any plan.

For the compute and volume layer, we don't name the underlying provider for a specific service. The authoritative list is our published subprocessor list, along with the DPA and regions documentation, which you can keep for your records.


Status changed to Awaiting User Response Railway • about 1 month ago


Railway

Volume backups (scheduled and on-demand) are stored in the region the service deploys to, so a volume in EU West (Amsterdam, Netherlands) is backed up in EU West. The Postgres point-in-time recovery archive (WAL segments and base backups) is also stored in the deploy region, in a Railway storage bucket. Bucket storage runs on Tigris, so for an EU West service that archive sits on Tigris in EU West. Neither of those native features copies data to a US region. Two caveats belong in your assessment. First, all application, deploy, build, and HTTP logs are stored in US West regardless of deploy region, so anything your Postgres service writes to its logs leaves the EU. Second, we keep separate internal platform backups for disaster recovery, and the facts above are not a location inventory for those, so we can't state that no copy of any data ever exists outside the EU. We also don't offer a contractual EU-only processing commitment on any plan. For the compute and volume layer, we don't name the underlying provider for a specific service. The authoritative list is our published [subprocessor list](https://trust.railway.com/item/subprocessors), along with the [DPA](https://railway.com/legal/dpa) and [regions documentation](https://docs.railway.com/deployments/regions), which you can keep for your records.

Anonymous
OP

a month ago

  1. Volume backups for this service: in which region are they

    physically stored?

  2. The point-in-time recovery archive is described as a bucket

    written by pgBackRest. In which region is that bucket located?

    We understand this is a separate mechanism from volume snapshots

    and may have a different answer.

  3. Uploads from the Postgres service to the PITR bucket are billed

    as service network egress. Does that indicate the bucket is

    outside the service's region or private network, or is it a

    billing convention only?

  4. Is either store replicated to, cached in, or processed in a

    United States region at any point?

Please answer as Railway rather than via the support assistant — we

need a position we can retain for a regulatory filing.


Status changed to Awaiting Railway Response Railway • about 1 month ago


Railway
BOT

a month ago

Volume backups, both scheduled and on-demand, are stored in the region the service deploys to. For a service in EU West (Amsterdam, Netherlands), the backups are stored in EU West.

The Postgres point-in-time recovery archive is the WAL segments and base backups that pgBackRest writes to the service's Railway storage bucket. That bucket is also in the deploy region, so for an EU West service it is in EU West. Railway storage buckets run on Tigris.

The egress charge is a billing rule. Every upload from a service to a bucket counts as service network egress, so the charge is not a sign that the bucket is in another region. Uploads are compressed first and billed on the compressed size, and downloads from the bucket during a restore are not charged.

Neither of these native backup features places its data in a US region. Two points from our earlier reply still apply to your assessment. First, all logs are stored in US West whatever the deploy region, so anything the Postgres service writes to its logs is stored there. Second, these facts describe the native backup features only. They are not a location inventory for our separate internal disaster-recovery copies, and we do not offer a contractual EU-only processing commitment on any plan.

For your records, the documented sources are the point-in-time recovery docs, the subprocessor list and the DPA.


Status changed to Awaiting User Response Railway • about 1 month ago


Railway

Volume backups, both scheduled and on-demand, are stored in the region the service deploys to. For a service in EU West (Amsterdam, Netherlands), the backups are stored in EU West. The Postgres point-in-time recovery archive is the WAL segments and base backups that pgBackRest writes to the service's Railway storage bucket. That bucket is also in the deploy region, so for an EU West service it is in EU West. Railway storage buckets run on Tigris. The egress charge is a billing rule. Every upload from a service to a bucket counts as service network egress, so the charge is not a sign that the bucket is in another region. Uploads are compressed first and billed on the compressed size, and downloads from the bucket during a restore are not charged. Neither of these native backup features places its data in a US region. Two points from our earlier reply still apply to your assessment. First, all logs are stored in US West whatever the deploy region, so anything the Postgres service writes to its logs is stored there. Second, these facts describe the native backup features only. They are not a location inventory for our separate internal disaster-recovery copies, and we do not offer a contractual EU-only processing commitment on any plan. For your records, the documented sources are the [point-in-time recovery docs](https://docs.railway.com/volumes/point-in-time-recovery), the [subprocessor list](https://trust.railway.com/item/subprocessors) and the [DPA](https://railway.com/legal/dpa).

Anonymous
OP

a month ago

🙂 a human answer. please.


Status changed to Awaiting Railway Response Railway • about 1 month ago


a month ago

Here are our answers, in order.

  1. Scheduled and on-demand volume backups are stored in the service's deploy region. For EU West, that is Amsterdam, Netherlands.

  2. The Postgres point-in-time recovery archive (WAL segments and pgBackRest base backups) sits in the service's PITR storage bucket, which is also in the deploy region, so EU West. Our storage buckets run on Tigris.

  3. We need to correct our earlier reply: the egress charge is not purely a billing convention. Buckets are on the public network and are not currently reachable over the private network, so uploads from the service to the bucket travel over the public network and are billed as service egress (Storage Buckets Billing). This reflects the network path, not a different region.

  4. Neither native backup feature places data in a US region. Two caveats for your filing. All application, deploy, build and HTTP logs are stored in US West regardless of deploy region, so anything Postgres writes to its logs is stored there. These answers also cover the native backup features only, not our separate internal disaster-recovery copies, and we do not offer a contractual EU-only processing commitment on any plan.

You can keep this reply as our written position, together with the PITR docs, the subprocessor list and the DPA.


Status changed to Awaiting User Response brody • about 1 month ago


brody

Here are our answers, in order. 1. Scheduled and on-demand volume backups are stored in the service's deploy region. For EU West, that is Amsterdam, Netherlands. 2. The Postgres point-in-time recovery archive (WAL segments and pgBackRest base backups) sits in the service's PITR storage bucket, which is also in the deploy region, so EU West. Our storage buckets run on Tigris. 3. We need to correct our earlier reply: the egress charge is not purely a billing convention. Buckets are on the public network and are not currently reachable over the private network, so uploads from the service to the bucket travel over the public network and are billed as service egress ([Storage Buckets Billing](https://docs.railway.com/storage-buckets/billing)). This reflects the network path, not a different region. 4. Neither native backup feature places data in a US region. Two caveats for your filing. All application, deploy, build and HTTP logs are stored in US West regardless of deploy region, so anything Postgres writes to its logs is stored there. These answers also cover the native backup features only, not our separate internal disaster-recovery copies, and we do not offer a contractual EU-only processing commitment on any plan. You can keep this reply as our written position, together with the [PITR docs](https://docs.railway.com/volumes/point-in-time-recovery), the [subprocessor list](https://trust.railway.com/item/subprocessors) and the [DPA](https://railway.com/legal/dpa).

Anonymous
OP

a month ago

Two follow-ups on point 3, since the bucket is on the public network:

  1. Are pgBackRest's uploads to the PITR bucket encrypted in transit (TLS to the Tigris endpoint)?

  2. Is the PITR bucket encrypted at rest, and is pgBackRest's own repository encryption enabled for it? If it is, who holds the key — Railway or us?

Asking because this is full database content crossing a public network, and our data-protection assessment needs the technical measures stated rather than assumed.


Status changed to Awaiting Railway Response Railway • about 1 month ago


sam-a
EMPLOYEE

a month ago

Yes, pgBackRest's uploads to the PITR bucket use HTTPS, so the archive is encrypted in transit with TLS to the bucket endpoint. Our Trust Center at trust.railway.com also publishes an Encryption-in-transit control that requires TLS.

Yes, the PITR bucket is encrypted at rest. The Encryption-at-rest control on the Trust Center reads: "All data at rest within Railway's systems is encrypted using AES 256, providing robust protection against unauthorized access and ensuring data confidentiality." It covers storage buckets, PITR archives (WAL segments and base backups) and volume backups. We handle encryption and key management, none of it is exposed to or configurable from your service, and we do not offer customer-managed keys.

For your assessment, the public control page and the SOC 3 report on the Trust Center can be kept without signing in. The SOC 2 Type II report and the Encryption Policy are available there as well.


Status changed to Awaiting User Response sam-a • about 1 month ago


Railway
BOT

a month ago

This thread has been marked as solved automatically due to a lack of recent activity. Please re-open this thread or create a new one if you require further assistance. Thank you!

Status changed to Solved Railway • about 1 month ago


Welcome!

Sign in to your Railway account to join the conversation.

Loading...