Independent comparison · no paid rankings
Home / Blog / Managed PostgreSQL: what you actually delegate
Guide

Managed PostgreSQL: what you actually delegate

Managed Postgres promises peace of mind, but patching, backups, and HA must be read in the contract. Here is what leaves your plate—and what stays your problem.

4 min read Updated Jul 19, 2026

The team migrates to managed Postgres to stop managing the database. First incident: connections exhausted because each serverless worker opens ten pools—the provider caps at one hundred connections on the entry tier. Second surprise: restore to the day, not the minute—an RPO incompatible with an online shop.

Managed PostgreSQL delegates machine operations: OS, engine, infrastructure backups, sometimes automatic failover. It does not delegate schema design, slow queries, or poorly tested migrations. Read the contract like a minimalist DBA—not like a promise of total peace of mind.

What the provider typically handles

TaskOften includedVerify
OS + minor Postgres patchYesMaintenance window, major version upgrade
Automatic backupsYesRetention, PITR, tested restore
HA failoverTier dependentAnnounced RTO, sync/async replication
Infra monitoringCPU, disk, connectionsConfigurable alerts
Encryption at-restOftenCustomer-managed keys?
Vertical scalingConsoleDowntime or online resize

What stays with you

Even with the best managed offer, several responsibilities never leave your team:

  • Schema and migrations — DDL, rollback strategy, N/N+1 compatibility between app versions.
  • Query performance — EXPLAIN, indexes, vacuum tuning sometimes limited by the host.
  • Connections — pooler (PgBouncer, Supavisor) in the app or as a managed add-on.
  • Logical security — roles, GRANT, row-level security configuration.
  • Sensitive data — application encryption, anonymization, GDPR compliance.
  • Restore drill — the button exists in the console; knowing how to click under stress does not.

If your stack still inherits MySQL, compare with MySQL for a dynamic site before switching.

Read an offer without being fooled

Before signing, ask these questions—and demand documented answers, not marketing phrases:

  1. Documented RPO/RTO for restore—not vague "daily backups."
  2. Required extensions available on your tier?
  3. Connection limit vs your architecture (Redis, serverless functions, multiple workers).
  4. Egress if the app runs in another region or cloud.
  5. Lock-in — pg_dump export, standard format, exit timeline.
  6. EU region and compliant DPA (GDPR host questions).
Weak signalStrong signal
"Backup included" with no retentionDocumented 7–35 day PITR
No pooler offeredManaged pooler or PgBouncer documentation
Postgres version frozen for yearsPublished upgrade policy

Managed Postgres vs self-admin VPS

Managed wins if you have no DBA, HA is required, audits demand proven backups, or nobody wants to patch at 3 a.m.

VPS wins if you need exotic extensions, fine kernel tuning, minimal fixed cost, or strong internal Postgres skills.

A common hybrid: managed production + monthly local restore drill.

The climax: the restore button is not a recovery plan

Require a quarterly restore exercise—even on a managed tier. On incident day, you will thank yourself for timing the procedure cold.

Decide and move forward without blind spots

Before choosing managed Postgres, align the contract with real usage:

  1. Set business RPO/RTO—how many minutes of data loss and hours of downtime are acceptable.
  2. Compare two or three European offers on extensions, connections, and point-in-time recovery.
  3. Add a pooler from day one, not after the first connection saturation.
  4. Automate migrations in your CI pipeline with review and documented rollback.
  5. Schedule a quarterly restore drill and record the actual elapsed time.

Browse our directory and comparator to filter hosts offering managed databases in your region.

Frequently asked questions

Does managed Postgres include application migrations?

No. The provider maintains the instance; you manage DDL and migrations through your code. A bad migration remains your responsibility.

Are auto backups enough?

Only if contract RPO/RTO match needs and you tested a real restore—see restore testing.

Are PostgreSQL extensions always available?

No. Each host publishes an allowlist. Verify PostGIS, pgvector, or TimescaleDB before locking a feature in.

RDS vs Scaleway vs OVH?

Different scope—compare pooler, regions, egress, and UI on an identical scenario.


Managed PostgreSQL: you delegate the database datacenter—not responsibility for your data or queries.

Compare European hosts

Filter by compliance, location and use case — then open the sheets to verify the real scope.

Browse the directory
Blog

Related reading

All articles →