Liquibase

Consider a scenario, currently, your application is using a Relational database, but management has decided to migrate to a NoSQL database, now All DDL and DML queries of all the database tables have to be written and run on a new NoSQL database.

or another scenario, with each new feature release, some new tables are created or changes in existing tables are required. the only way to make changes in the production database is to write DDL and DML SQL queries and provide them to the database team, who will run on production, and then only the application can be deployed.

If application deployment fails, then the application state needs to be reverted back to the previous state. reverting application code is very easy with the help of git commit or branch, but reverting the database to the previous state is very tricky and error-prone, each DDL and DML query has to be observed and a new revert query must be written.

What is Liquibase

Liquibase, an open-source library, tries to address all these issues with the git commit way. each database change state is saved in a change log file with a version. 
changelog supports SQL JSON and YAML as well but XML is more preferred due to readability.

Each change log version stores the state of the database including naming conventions, grants & revokes, and rollback scripts, with the help of the changelog version id, the database can be easily reverted back to the previous state.
Liquibase compares the current state of the database with the changelog file and only makes changes that are not available in the database.
change log file keeps reverting and rolling back for each database state change log, which can be executed automatically if the update fails or manually as well as per requirement.
These changelog files are database agnostic, hence changes made for a type of database can easily be migrated to other databases smoothly. Liquibase supports most of the popular databases including MySQL, Oracle, DB2, and PostgreSQL.