pgloader

pgloader is a self-hosted migration utility for bringing supported databases and flat files into PostgreSQL. It is best for teams that can run their own tooling and need schema conversion with parallel loading. Its focused PostgreSQL workflow is compelling for supported sources, but version status, Oracle coverage, and SQL-object limitations deserve careful review before adoption.

Overview

pgloader moves data through five stages: fetching the source catalog, preparing the target, running parallel COPY operations, creating indexes in parallel, and completing the schema. That sequence suits one-time or project-based migrations where throughput and repeatable conversion rules matter. Discovery and assessment plus dependency mapping help plan a move, while the absence of continuous replication makes it unsuitable as an ongoing synchronization service.

Sources include MySQL, MariaDB, MS SQL Server, SQLite, CSV, fixed-width files, and DBF. Oracle migration is not built yet, so Oracle teams need another route. Dimitri Fontaine has authored and maintained the project since 2005.

Key features

Source-aware conversion

The MySQL path translates enum values into PostgreSQL types, turns TINYINT(1) into boolean, handles zero dates, and can read credentials from .my.cnf. These choices address common MySQL conversion friction without requiring every adjustment to be hand-written. For MS SQL Server, Azure AD authentication, sequence migration, and multi-column foreign-key support cover important enterprise schema details. They help SQL Server projects in particular, although they do not amount to a universal SQL translator.

Parallel loading and file handling

Parallel data copying and index creation can shorten migration stages on suitable workloads. File imports are more forgiving than a simple all-or-nothing load: when rows cause a COPY transaction to fail, pgloader separates rejected rows, retries, and continues loading accepted data. That behavior is valuable for imperfect CSV, fixed-width, or DBF feeds where isolating bad records is preferable to restarting the entire import.

Custom rules and boundaries

A .load file accepts custom casting rules, giving operators control when built-in conversions do not match application expectations. The larger limitation is architectural: the roadmap says pgloader works from catalog representations and does not currently parse SQL object definitions. View definitions and stored procedures therefore need SQL-translation support outside this tool, adding work for applications that depend heavily on programmable database objects.

Runtime and release position

Version 4 is distributed as a self-contained JAR and requires Java 21 or later. It is described as a pre-release, while version 3 in Common Lisp remains the stable, packaged release. Teams prioritizing a conservative production baseline should weigh the newer runtime against the maturity of the stable line.

Pricing

pgloader uses a freemium support model. The Community tier costs nothing and is intended for evaluation or non-critical use, with best-effort help through GitHub. It is the sensible starting point for teams that can manage migrations independently and accept community-level response.

The paid tiers buy response, patch handling, or development commitment rather than a hosted service. None changes the self-hosted deployment model, and no continuous-replication capability is included.

Platforms

pgloader runs on Linux and is self-hosted. Version 4 requires Java 21 or later; the stable packaged v3 line uses Common Lisp. Teams must therefore own the runtime, deployment, and operational responsibility rather than relying on a vendor-managed endpoint.

Who it's for

Choose pgloader when a Linux-capable team is moving MySQL, MariaDB, SQL Server, SQLite, or supported file formats into PostgreSQL and wants parallel loading plus configurable casting. It also suits organizations that need discovery, assessment, and dependency mapping before a migration. Look elsewhere when Oracle is the source, when views and stored procedures must be translated automatically, or when ongoing replication is a requirement.

Pros and cons

• Pros: Broad coverage across several databases and flat-file formats; parallel data and index stages; useful MySQL and SQL Server-specific conversions; retry-and-isolate behavior for rejected file rows; custom casting through .load files; self-hosted operation.
• Cons: No Oracle source support; no continuous replication; SQL object definitions, views, and stored procedures are outside current parsing capability; v4 is pre-release and requires Java 21; Community support is best effort.

Alternatives

For a broader cloud migration program, consider Cloud Migration Tools. AWS Transform MGN (formerly AWS Application Migration Service) is a paid service with a free period of up to 2,160 hours per source server, focused on AWS replication infrastructure rather than a self-hosted PostgreSQL loader. Google Cloud Migration Center is free for cost estimates and assessments, though other Google Cloud products can incur charges. Azure Application Insights uses usage-based Azure Monitor Logs pricing with the first 5 GB per month per billing account free, making it an observability-oriented alternative rather than a direct loader.

Google Compute Engine offers a freemium compute platform with a free tier and trial, useful when you need infrastructure to run migration workloads. Amazon Database Migration Service is a paid AWS option with free-tier credits valid for up to 12 months. AWS Schema Conversion Tool is a free desktop-oriented schema conversion utility for supported Windows and Linux distributions. Google Migrate to Containers is free for migration work but targets containerizing workloads. For Oracle-to-PostgreSQL projects, Ora2Pg is a free GPL v3 alternative.

Verdict

pgloader is a strong choice for technically capable teams performing controlled migrations into PostgreSQL from its supported sources, especially when parallel loading and source-specific casting reduce conversion effort. Choose another product when you need Oracle support, translation of views and stored procedures, managed cloud operations, or continuous replication.