Schema migrations you can prove.

MigAce migrates a database with real data and then proves the result is identical to the schema your application expects, down to the last constraint. No drift, no surprises in production.

See how it works Star on GitHub

# migrations/2026-11-11-add-customer-email.yaml
id: add-customer-email
parts:
  - sql: ALTER TABLE customer ADD COLUMN email text
  - data: UPDATE customer SET email = legacy_mail
  - sql: ALTER TABLE customer ALTER email SET NOT NULL

$ ./gradlew migaceVerify
✔ schema A  generated from Hibernate model
✔ schema B  restored from dump + 214 snippets applied
✔ A ≡ B     0 differences  (digest 9f3c…e1a7)
✔ HA check  compatible with release 4.12

The proof loop

Every build runs the same loop. If a migration doesn't produce exactly the expected schema on a database that already contains data, the build fails, long before production.

01 · TARGETGenerate schema A from your Hibernate model.

02 · RESTORELoad an older production-like dump with data into schema B.

03 · MIGRATEApply all pending snippets to B, part by part, fully logged.

04 · COMPAREDiff A and B: tables, columns, constraints, indexes, routines.

05 · SEALStore the schema digest. The app refuses to start on a mismatch.

Built for real enterprise databases

Born in a long-running Java EE project and rebuilt from scratch for PostgreSQL.

🎯

Zero-drift guarantee

Production schema matches the Hibernate-generated DDL exactly, verified on every build.

🔁

Re-runnable snippets

YAML snippets made of small parts. Each part is logged and skipped once applied, so a failed run simply resumes.

📦

Migrations on real data

Adding NOT NULL or unique constraints is tested against data, not just an empty schema.

🧮

Data migrations

First-class data and repair snippets, separate from schema changes and fully traceable.

🚦

HA & rolling updates

Classifies every change as compatible or breaking, so older app versions keep running during a Kubernetes rollout.

🔐

Schema digest at startup

The app checks a hash of the live schema and refuses to start against an unexpected database.

🧪

Test data generation

Generate realistic data for every new table and column so each snippet is exercised.

⚙️

Gradle-native

A Kotlin Gradle plugin with Testcontainers built in. One task: migaceVerify.

☸️

Runs where you deploy

Execute from an init container or job, or export a plain SQL script for your DBAs.

How MigAce compares

Indicative comparison for planned features. MigAce is not released yet.

Capability	MigAce	Typical versioned tools
Proves migrated schema = ORM model	Yes	Manual / partial
Tests migrations against existing data	Built in	Up to you
Resumable, part-level execution log	Yes	Per-file
Rolling-update compatibility check	Yes	Rare
Schema digest check at app startup	Yes	Rare

Get started (preview)

The plugin is not published yet. This is what using it will look like.

// build.gradle.kts
plugins {
    id("org.migace") version "0.1.0"
}

migace {
    database = postgres("17")
    snapshot = file("db/baseline.dump")
    snippets = file("db/migrations")
}