Migrate Azure Database for PostgreSQL to Hetzner Cloud – pg_dump/restore Guide

Azure costs exploding? Flexible Server too expensive? We migrate your PostgreSQL databases from Azure to Hetzner Cloud – with minimal risk and a clear cutover plan. 👉 Azure Migration Overview | Request Free Analysis

When teams migrate out of Azure, the database is often the most critical component: costs and lock-in play a role, but in day-to-day operations what matters most is predictability, data integrity, and a clean cutover. For Azure Database for PostgreSQL (Flexible Server), Microsoft describes Dump & Restore as the standard approach and recommends Directory format + parallelization for large databases (Microsoft Learn).

For the target: In Hetzner Cloud there's no "Azure-like Managed PostgreSQL" – you build the target as a VM + self-installed PostgreSQL (and handle backups/monitoring yourself). For VM backups, you can enable Daily Backups at Hetzner (7 slots, rotation) (Hetzner Docs).

Table of Contents

• Recommendation and Target Architecture
• Pre-Migration Checklist
• Prepare Target System: Hetzner VM and PostgreSQL
• Path 1: Dump and Restore (Offline)
• Validation After Migration
• Path 2: Logical Replication (Low Downtime)
• Common Pitfalls
• Cutover Checklist
• Our Approach at WZ-IT
• Related Guides

Recommendation and Target Architecture

Default Recommendation: pg_dump/pg_restore

If you're using Azure and a maintenance window (downtime) is acceptable:

• The approach is proven, stable, and transparent – Microsoft provides very specific best practices (Microsoft Learn)
• Complex replication or DMS setups are eliminated
• For databases up to ~100-500 GB, a maintenance window is realistically plannable

Low-Downtime (Path 2) is worth it when:

• your maintenance window is too small
• the DB is so large that restore takes too long
• or you can't cleanly "freeze" writes

Pre-Migration Checklist

Before you dump anything:

1. Check Postgres Major Version (Azure → Target should be compatible; ideally: same major version)
2. Inventory Extensions

   SELECT extname, extversion FROM pg_extension ORDER BY 1;
   

3. Roles/Privileges: Who needs what rights after the move?
4. Size + Downtime Tolerance: determines dump format, parallelism, cutover window
5. Network Access: Target VM must be reachable from Azure

Prepare Target System: Hetzner VM and PostgreSQL

Create VM and Storage

1. Create VM in Hetzner Cloud
   • Ubuntu LTS (e.g., 22.04) or Debian
   • Sufficient RAM/CPU for your DB load
   • Separate Volume for /var/lib/postgresql (growth/backups)
   • Private Network / Firewall: Port 5432 only from admin/migration IPs

Install PostgreSQL

sudo apt update
sudo apt install -y postgresql postgresql-contrib
sudo systemctl enable --now postgresql

Create database and user:

sudo -u postgres createuser --pwprompt appuser
sudo -u postgres createdb -O appuser appdb

Enable Backups

At Hetzner you can enable Daily Backups for Cloud servers (7 slots, oldest gets overwritten) (Hetzner Docs). This doesn't replace DB logical backups, but it's a good safety net.

Path 1: Dump and Restore (Offline)

Microsoft describes the procedure for Azure Database for PostgreSQL Flexible Server explicitly including role export, dump options, and restore; the Directory format is recommended because it enables parallelization (Microsoft Learn).

Workflow (High-Level)

1. Start maintenance window – App in Maintenance / Read-Only
2. Export roles (separately with pg_dumpall)
3. Create dump (Directory format, parallel)
4. Transfer dump to Hetzner VM
5. Restore on target PostgreSQL (parallel)
6. Fixups – Ownership, Grants, Sequences, Stats
7. Validation – Counts, spot checks, smoke tests
8. Cutover – Switch app connection string
9. Monitoring – Keep Azure DB initially (rollback window)

Step 1: Export Roles/Users

Microsoft explicitly lists the pg_dumpall -r pattern as a step for roles/users (Microsoft Learn).

export PGPASSWORD="<AZURE_PASSWORD>"

pg_dumpall -h <AZURE_HOST> -p 5432 -U <AZURE_ADMIN_USER> -r > roles.sql

Tip: Review the roles file later (don't execute blindly) to avoid unwanted owner/privileges.

Step 2: Dump Database (parallel)

Directory Format (recommended for large DBs):

export PGPASSWORD="<AZURE_PASSWORD>"

pg_dump \
  -h <AZURE_HOST> -p 5432 -U <AZURE_ADMIN_USER> -d <DBNAME> \
  -Fd -j 8 \
  -f dumpdir/

Custom Format (for small/medium DBs):

pg_dump \
  -h <AZURE_HOST> -p 5432 -U <AZURE_ADMIN_USER> -d <DBNAME> \
  -Fc -Z 6 \
  -f db.dump

Microsoft has a dedicated article "Best practices for pg_dump and pg_restore" with parallelism and VM considerations (Microsoft Learn).

Step 3: Transfer to Hetzner VM

rsync -avP dumpdir/ root@<HETZNER_VM_IP>:/root/dumpdir/
rsync -avP roles.sql root@<HETZNER_VM_IP>:/root/

Step 4: Restore on Hetzner VM

Create database:

export PGPASSWORD="<HETZNER_PASSWORD>"
createdb -h <HETZNER_VM_IP> -p 5432 -U <HETZNER_USER> <DBNAME>

Import roles (with care):

psql -h <HETZNER_VM_IP> -p 5432 -U <HETZNER_USER> -f /root/roles.sql

Restore (Directory, parallel):

pg_restore \
  -h <HETZNER_VM_IP> -p 5432 -U <HETZNER_USER> -d <DBNAME> \
  --no-owner --no-privileges \
  -j 8 \
  /root/dumpdir/

Why --no-owner --no-privileges? When switching providers, you want to apply roles/grants deliberately to your target model, rather than importing Azure defaults.

Update statistics:

ANALYZE;

Validation After Migration

Hard Checks (Data)

• Rowcounts for critical tables
• Sum checks (e.g., invoice amounts)
• Spot checks (50-100 IDs) across tables

SELECT 'users'  AS t, count(*) FROM public.users
UNION ALL
SELECT 'orders' AS t, count(*) FROM public.orders;

Smoke Tests (Application)

• Login works
• 2-3 most important business processes (Create/Update/Payment)
• Background Jobs / Cron
• Test write paths (Insert/Update/Delete)

Path 2: Logical Replication (Low Downtime)

Azure Flexible Server supports native logical replication and logical decoding; Microsoft explicitly describes Publisher/Subscriber models (Microsoft Learn). PostgreSQL explains: Each subscription uses a replication slot (PostgreSQL Docs).

High-Level Runbook:

1. Target VM ready, schema prepared
2. On Azure: Create publication
3. On Hetzner: Subscription (Initial Sync + Streaming)
4. Monitor lag
5. Cutover: short write-freeze → Catch-up → Switch app → Clean up subscription/slots

Logical Replication is powerful – but has more moving parts (network, slots, lag, maintenance). Therefore, with acceptable downtime: Dump/Restore remains the default.

Common Pitfalls

Cutover Checklist

T-7 to T-2 Days

• Hetzner VM + Volume + Firewall ready
• PostgreSQL installed, Monitoring/Logs ok
• Hetzner Server Backups enabled (7 slots)
• Extensions & versions checked
• Test run (small DB or snapshot) successful

T-24h

• Maintenance window communicated
• Rollback plan documented (Connection string back)
• Smoke test scenarios defined

Cutover

• App Read-Only / Maintenance
• pg_dumpall -r + pg_dump + Transfer + pg_restore
• Validation: Counts + spot checks + smoke tests
• App switched to Hetzner DB
• Monitoring (Errors, DB connections, latency)

T+24h

• Keep Azure DB as fallback
• Monitor performance & errors
• Decommission Azure DB when stable

Our Approach at WZ-IT

Our standard flow for Azure PostgreSQL → Hetzner:

1. Assessment: DB size, extensions, change rate, downtime budget
2. Target System: Provision Hetzner VM, install PostgreSQL, firewall/backup
3. Test Migration: Test dump on staging DB, validation, performance check
4. Cutover Planning: Maintenance window, rollback plan, communication
5. Migration: Dump → Transfer → Restore → Validation → Cutover
6. Monitoring: Error rates, latencies, app logs, backup verification

👉 We create a concrete cutover runbook with all commands and checklists for you.

Related Guides

You might also be interested in these guides:

• Migrate Azure Blob Storage to Hetzner Object Storage – AzCopy & rclone Guide
• Azure to Hetzner Migration – Complete Overview – All Azure services at a glance
• Cloud Migration Hub – All provider migrations at a glance
• Migrate AWS RDS/Aurora PostgreSQL to Hetzner – pg_dump/restore Guide
• Hetzner Expertise – Our Hetzner services overview

Ready for migration? At WZ-IT, we handle analysis, migration, and operation of your database infrastructure – from Azure to cost-effective European alternatives.

👉 Request Free Cost Analysis