Database Migration: Step-by-Step Guide, Tools & Checklist

What is database migration?

In operations, database migration (often shortened to db migration) means moving a database from one place to another: to a new server or data centre, to a newer engine version, to a managed cloud service such as Amazon RDS or Azure SQL, or to a different engine altogether, for example SQL Server or MySQL to PostgreSQL, or Oracle to PostgreSQL. The first three are homogeneous migrations (same engine); the last is heterogeneous and adds schema and code conversion.

In application development the same phrase means something else: a versioned script that changes the schema (add a column, create an index) and is applied in order to every environment. Both meanings are covered here, the first in depth, the second in the last section.

The database migration process, step by step

The six steps below apply to any engine or platform. Our database migration checklist breaks each one into individual checks you can tick off.

Assess before you choose a tool

The assessment decides everything else. Record the source engine and edition (AWS DMS, for instance, cannot read changes from SQL Server Express), total size and the largest tables, data types that have no direct equivalent on the target, stored procedures and triggers, and every application, report and job that connects. Engine-specific tools produce an assessment report for you: DMS Schema Conversion and SSMA both estimate how much converts automatically and what needs manual work.

Convert the schema

For a same-engine move, the schema travels with the dump or backup. For a heterogeneous move, convert it with a tool built for the pair: DMS Schema Conversion for several sources into Amazon RDS, Aurora or Redshift; SSMA for Oracle, MySQL, Access, Db2 and SAP ASE into SQL Server or Azure SQL; pgloader for MySQL, SQL Server and SQLite into PostgreSQL; Ora2Pg for Oracle (and MySQL) into PostgreSQL. Expect manual work on procedural code: T-SQL or PL/SQL procedures rarely convert line for line.

Move the data

Load the data with constraints and secondary indexes off where the tool allows it, and build them afterwards; AWS gives this advice for large DMS full loads, and pg_restore's parallel jobs speed up index and constraint creation. For an online migration, start change capture before or with the bulk copy so nothing written during the copy is lost.

Validate, cut over and plan the rollback

Compare the two sides table by table (see the worked example below), then run the application against the target. At cutover, stop writes to the source, let the last changes apply, set each sequence or identity on the target past the highest value copied, and switch the connection string. Keep the source unchanged and readable until you are confident; the rollback plan is simply pointing the application back, so agree beforehand which failures would trigger it and how long the window stays open.

Offline or online migration: choosing a method

Offline (dump and restore). Stop the application, take a consistent export, restore it on the target, validate, switch. pg_dump makes consistent exports even while the database is in use, and mysqldump's --single-transaction option dumps InnoDB tables inside one transaction. Downtime equals export plus transfer plus restore plus checks, so it suits databases that copy within your maintenance window.

Online (replicate, then switch). Copy the data while the source stays live, keep applying changes, and switch in a short window. Options include a managed service (AWS DMS full load plus CDC, Azure Database Migration Service online mode for SQL Managed Instance) or the engine's own replication for same-engine moves. It costs more effort and needs keys on every table, but downtime shrinks to the final switch.

Database migration tools compared

Choose by source, target and whether you need ongoing replication. All of the tools below are free to download or use except the managed services, which bill for the capacity they run on.

Worked example: an offline PostgreSQL move with validation

The simplest real migration is a same-engine move, for example a PostgreSQL database moving to a new server or to a managed service. The commands follow the examples in the PostgreSQL documentation: dump to a custom-format archive, create an empty database from template0, and restore with parallel jobs. --no-owner is useful when the target uses different role names, as managed services usually do. For MySQL the equivalent is mysqldump, and for SQL Server a native backup and restore; the pg_dump and mysqldump guides in the related pages cover the options in full.

Run the validation query on both sides and compare the output. Matching counts, sums and key ranges do not prove every value is identical, but they catch missing batches, truncated loads and duplicate inserts quickly; add a sample of rows compared column by column for critical tables.

# 1. On the source (application stopped or read-only)
pg_dump -h <source-host> -U <user> -Fc mydb > mydb.dump

# 2. On the target: empty database, then a parallel restore
createdb -h <target-host> -U <user> -T template0 mydb
pg_restore -h <target-host> -U <user> -d mydb -j 4 --no-owner mydb.dump

SELECT COUNT(*)          AS row_count,
       SUM(total_amount) AS total_amount,
       MIN(order_id)     AS min_id,
       MAX(order_id)     AS max_id
FROM orders;

Common migration scenarios

• SQL Server to PostgreSQL: type mapping, T-SQL to PL/pgSQL, and the choice between pgloader, AWS DMS with Schema Conversion, and Babelfish for Aurora PostgreSQL. See the SQL Server to PostgreSQL migration guide.
• MySQL to PostgreSQL: pgloader in one command, AUTO_INCREMENT to identity columns, zero dates and case sensitivity. See the MySQL to PostgreSQL migration guide.
• Oracle, MySQL or Access to SQL Server or Azure SQL: SSMA. See the SQL Server Migration Assistant guide.
• Oracle database migration from on-premises to the AWS cloud: DMS Schema Conversion converts Oracle schemas for Aurora or RDS for PostgreSQL or MySQL, and AWS DMS moves the data with optional CDC. Ora2Pg is the open-source route to PostgreSQL.
• Azure database migration: Azure Database Migration Service for SQL Server into Azure SQL Database or Managed Instance; SSMA when the source is not SQL Server.
• DynamoDB migration: Amazon DynamoDB is one of AWS DMS's documented targets, so relational data can be loaded into it with a DMS task; plan the key design first, because tables do not map one to one.

Schema migrations: the other meaning of "database migration"

Development teams use "migrations" for the scripts that evolve a schema over time. Flyway, for example, describes migrations as SQL scripts that capture schema and data changes, are kept in version control and are run in the same order in every environment. Its migrate command compares the scripts with a schema history table in the database and applies only those not yet run, so a database at version 5 with scripts up to version 9 receives 6, 7, 8 and 9. Versioned migrations can have matching undo migrations, and repeatable migrations are re-applied when they change.

Frameworks in most languages ship their own migration runners, including ones for document databases such as MongoDB; the idea is the same: small, ordered, version-controlled changes. Use schema migrations for day-to-day changes, and the process above when the whole database moves.

Next steps

Before a heterogeneous migration, read the engine comparison so you know which features will need rework: SQL Server vs PostgreSQL, MySQL vs PostgreSQL or PostgreSQL vs Oracle. If you are new to SQL or need a refresher on the statements used to check a migration, start with the SQL beginner course.

Frequently asked questions

Sources

• PostgreSQL documentation: pg_dump
• PostgreSQL documentation: pg_restore
• PostgreSQL versioning policy
• MySQL 8.4 Reference Manual: mysqldump
• Microsoft Learn: bcp utility
• Microsoft Learn: Back up and restore of SQL Server databases
• Microsoft Learn: SQL Server Migration Assistant
• Microsoft Learn: What is Azure Database Migration Service?
• AWS DMS User Guide: What is AWS DMS?
• AWS DMS User Guide: Best practices
• AWS DMS User Guide: Targets for data migration
• AWS DMS User Guide: DMS Schema Conversion
• pgloader documentation
• pgloader releases (GitHub)
• Ora2Pg home page
• Ora2Pg documentation
• Redgate Flyway documentation: Migrations

Checked 8 October 2026.

How we research guides: our editorial method. We link only to official downloads and never host installers.