A changelog is the root file that contains or references all database changes for your project. It acts as a ledger of all modifications that should be applied to bring a database to a particular state.

A changeset is an individual unit of change within a changelog. Each changeset represents a single logical database change and has a unique identifier.

<!-- This is a changelog file -->
<databaseChangeLog>
    <!-- This is a changeset -->
    <changeSet id="1" author="john.doe">
        <createTable tableName="users">
            <column name="id" type="int" autoIncrement="true">
                <constraints primaryKey="true"/>
            </column>
            <column name="username" type="varchar(50)"/>
        </createTable>
    </changeSet>
    
    <!-- Another changeset -->
    <changeSet id="2" author="jane.smith">
        <addColumn tableName="users">
            <column name="email" type="varchar(100)"/>
        </addColumn>
    </changeSet>
</databaseChangeLog>

Key characteristics of changesets:

• Must have unique ID and author combination
• Are atomic (all-or-nothing execution)
• Can only be run once per database (unless marked as runAlways)

Liquibase supports four main formats for writing changelogs:

1. XML - The original and most feature-complete format
2. YAML - Human-readable alternative to XML
3. JSON - Useful for programmatic generation
4. SQL - Direct SQL statements with Liquibase formatting comments

Example of the same changeset in different formats:

XML:

<changeSet id="1" author="developer">
    <createTable tableName="products">
        <column name="id" type="int" autoIncrement="true">
            <constraints primaryKey="true"/>
        </column>
    </createTable>
</changeSet>

YAML:

changeSet:
  id: 1
  author: developer
  changes:
    - createTable:
        tableName: products
        columns:
          - column:
              name: id
              type: int
              autoIncrement: true
              constraints:
                primaryKey: true

SQL:

--liquibase formatted sql
--changeset developer:1
CREATE TABLE products (
    id INT AUTO_INCREMENT PRIMARY KEY
);

Liquibase uses two system tables to track applied changes:

1. DATABASECHANGELOG: Records each changeset that has been executed

   • ID: Changeset identifier
   • AUTHOR: Changeset author
   • FILENAME: Source changelog file
   • DATEEXECUTED: When the changeset was run
   • MD5SUM: Checksum of the changeset content
   • EXECTYPE: How it was executed (EXECUTED, FAILED, SKIPPED, etc.)

2. DATABASECHANGELOGLOCK: Prevents multiple instances from modifying the database simultaneously

   • ID: Lock identifier
   • LOCKED: Boolean indicating if database is locked
   • LOCKGRANTED: Timestamp when lock was acquired
   • LOCKEDBY: Information about who holds the lock

When Liquibase runs, it:

1. Checks the lock table to ensure exclusive access
2. Compares changelog contents with DATABASECHANGELOG records
3. Executes only new/unrun changesets
4. Updates tracking tables with execution results