🧪 Spring Boot Integration Tests & Database Lifecycle

This document explains how Spring Boot manages the database lifecycle during integration tests — specifically when the database is created, when it's dropped, and how isolation is achieved.

> 🛢️ Database Creation and Dropping

>> When is the database created?

---

When using an in-memory database like H2 and spring.jpa.hibernate.ddl-auto=create-drop:

• ✅ The database schema is created once, when the application context is first started.
• This usually happens once per test class, if you're using @SpringBootTest.
• The same context (and DB schema) is reused across test methods in the same class.
• ❌ The database is not recreated before each test method.

>> When is the database dropped?

---

• With ddl-auto=create-drop, the schema is dropped automatically when the Spring context shuts down, i.e.:
• After the test class finishes (if context was specific to that class)
• Or at the end of the full test suite (if context was reused)

> ♻️ Transactional Rollback and Isolation

>> What does @Transactional do?

---

When a test class or test method is annotated with @Transactional:

• Each test method runs in its own transaction
• After the test method completes, the transaction is rolled back
• This leaves the database in a clean state between tests — without recreating the schema

>> Example

---

@SpringBootTest
@Transactional
class MyIntegrationTest {

    @Autowired
    UserRepository userRepository;

    @Test
    void testCreateUser() {
        userRepository.save(new User("Alice"));
        // Rolled back after the test method finishes
    }
}

> 🔎 Controlling DB and Context Scope

Situation	Context Reused?	DB Recreated?	Notes
Multiple tests in same class	✅ Yes	❌ No	Same context, same schema
Multiple test classes (default)	✅ Yes	❌ No	Spring optimizes by reusing context
With @DirtiesContext	❌ No	✅ Yes	Forces full context + DB reset
With @Transactional	✅ Yes	❌ No	Rollback used instead of full reset

> 👍 Best Practices for Integration Tests

✅ Use @Transactional to ensure clean state without heavy DB recreation
✅ Use ddl-auto=create-drop for in-memory DBs (e.g. H2) to keep schema fresh
✅ Use @DirtiesContext only when you need full context isolation
✅ Use fixture loaders (e.g. JSON or SQL) to control test data explicitly
⚠️ Avoid ddl-auto=update in tests — unpredictable and risky
⚠️ Avoid mixing stateful and stateless test strategies in the same class

> ⚙️ Common Configuration Example

spring:
  datasource:
    url: jdbc:h2:mem:testdb;DB_CLOSE_DELAY=-1
    driver-class-name: org.h2.Driver
  jpa:
    hibernate:
      ddl-auto: create-drop
  sql:
    init:
      mode: embedded

📌 Summary

• Spring Boot creates the DB once per context, usually once per test class.
• Schema is created at context startup, dropped at context shutdown.
• Use @Transactional for method-level isolation.
• Use @DirtiesContext if you need class-level isolation with fresh schema.