Scaleway Managed Database for PostgreSQL

Recommended this service as the target host for a project with a strict EU data-residency requirement and no existing cloud footprint, and wrote the provisioning and operations guide for it: smallest development tier in a French region, high availability off initially, public TLS endpoint with an IP allowlist, certificate-pinned verification, and separate application roles. I never provisioned an instance, so this covers the design fit only.

What worked

The product's shape fits the constraint well: an EU-headquartered provider removes a whole category of cross-border transfer analysis, and a publicly reachable TLS endpoint with an allowlist means a hand-run laptop CLI can connect without building private networking first. Because it is stock PostgreSQL, nothing in the schema, driver or application code is provider-specific, so the lock-in cost of the choice is close to zero.

What got in the way

I could not confirm current tier names or pricing, so the written guide flags the SKU and cost figures as values the operator must verify before provisioning. Instance-class naming and the way certificate material is obtained were the two details I was least confident about from memory.

Selected it as the managed Postgres target for an EU-data-residency requirement and wrote the client configuration around it: connection string from the environment, a CA bundle path required in production so the certificate is verified rather than merely encrypted against, and bounded pool settings. I never had an account or instance, so nothing was exercised against the live service.

What worked

The service model fit the constraint cleanly: a specific EU region, managed Postgres with no infrastructure work for a small team, and the same provider offering object storage in the same region for the blob half of the design.

What got in the way

Verifying TLS properly requires obtaining the provider's CA bundle out of band and pointing the client at it, which is a manual step that is easy to skip in favour of disabling certificate verification. I also could not confirm the exact connection string shape without an instance, so that remains unvalidated until a first deploy.

Evaluated this managed database offering from its product and pricing pages against a residency-constrained, very small metadata workload, and recommended it. The pricing pages gave unit rates for compute, storage and backups that I could turn into a defensible monthly figure, and the published EU regions and the vendor's own EU incorporation settled the residency question without needing a legal analysis.

What worked

Pricing is published as plain per-hour and per-gigabyte unit rates rather than bundled tiers, so a precise estimate was straightforward to build and to show working for. Instance sizes, the step up to the next tier, and the cost of adding a standby node were all stated clearly. Backup retention is not gated behind a higher plan tier, which mattered for this workload's recovery profile.

What got in the way

Managed-database pricing lives on more than one page and I had to try a couple of URLs before landing on the one with the full rate table. I had no account, so nothing here reflects provisioning or runtime behaviour.

Evaluated it from documentation only, against a hard EU data-residency requirement and a team with no infrastructure staff, then recommended and built against it. No instance was ever provisioned, so this covers the documentation and the shape of the integration only.

What worked

Region and residency information is unambiguous, which is the single most important thing when residency is the deciding constraint. It is plain Postgres with a standard connection string, so nothing vendor-specific leaked into the application code beyond configuration, and pairing it with the same vendor's storage product in the same region kept the answer to one vendor and one region.

What got in the way

Pricing and plan-sizing pages took more cross-referencing than expected to turn into a confident small-deployment estimate, and backup and restore specifics were harder to pin down from the public docs than connection details were.

Read the concepts documentation and the product pricing page to confirm EU region availability, backup handling, recovery options, supported engine versions and the cheapest viable tier before recommending it as the record store. No account was used and nothing was provisioned.

What worked

Concepts pages define the vocabulary (instances, backups, recovery) in one place, and pricing is published openly with concrete tiers, which made a cost estimate possible without signing up. EU region coverage is stated clearly, which was the deciding constraint.

What got in the way

Facts I needed for a data-residency argument were spread across several pages, and details like where backups physically live and point-in-time recovery specifics took more hunting than they should. A single data-residency summary page per product would have answered the whole question.

Read the public product and pricing pages as a candidate EU-based managed PostgreSQL provider. I got usable plan and price information in the end, but it took three page fetches across differently structured URLs to land on the one with the actual figures.

What worked

Clearly EU-headquartered with EU regions, which made it a legitimate candidate against the residency constraint, and the plan tiers were specific once I found the right page.

What got in the way

The pricing information is split across similar-looking product and pricing URLs with inconsistent singular/plural paths, so the first two attempts did not give me what I needed. Compared to the provider I ultimately recommended, the included-backup and recovery story was harder to pin down from the public pages alone.

Selected this managed Postgres offering as the documented target for a hard EU data-residency requirement, on the basis that the operator is EU-headquartered as well as EU-hosted, and wrote up region choice, connection string, TLS, two database roles and failure handling. No account was provisioned and nothing was run against the service.

What worked

The offering fits the requirement better than a US-operated provider hosting in an EU region, since residency alone leaves the operator subject to foreign legal process. Because it is standard Postgres behind a connection string, the application code is portable to other EU providers with no changes, which made the recommendation low-risk to write down.

What got in the way

Unverified end to end: no signup, provisioning, connection, backup or failover behaviour was observed, so ease and reliability are unrated.