AppSuite:DBMigration

From Open-Xchange
Revision as of 06:10, 29 August 2014 by Martin.schneider (talk | contribs)


This information is valid from 7.6.1 on.
How to write custom bundles to execute database migration statements.

Summary: With release 7.6.1 it is possible to change the database schema based on usage of the open source tool Liquibase. Currently only updating the configdb with the new mechanism is desired. This article gives a short introduction based on an existing sample bundle how to write custom database migration bundles and how to attach your custom statements to those given by Open-Xchange.

Please have a look at liquibase, its features and documentation before writing custom bundles: http://www.liquibase.org/documentation/ .

Prerequisite

This article is based on an existing sample bundle located in the public git repository backend-samples. Clone the repository by executing

git clone https://git.open-xchange.com/git/examples/backend-samples

The required bundle is named 'com.openexchange.sample.database.migration'.

Current restrictions

  • Database migrations can only be applied to the configdb!
  • Open-Xchange server currently uses liquibase version 3.0.7. All features provided with later versions cannot be used.

Bundle dependencies

To execute custom database migration for the configdb you have to track the service com.openexchange.database.migration.DBMigrationExecutorService (default provided within bundle com.openexchange.database.migration).

If you would like to have the Open-Xchange database migrations executed before your custom statements are used you even have to track the service com.openexchange.database.migration.ox.DBMigrationOXExcecutorService which is provided within the bundle com.openexchange.database.migration.ox

Liquibase

The following chapter describes in short the liquibase features and how they are used by Open-Xchange. For further information have a look at the official liquibase documentation: http://www.liquibase.org/documentation/

ChangeLog file / changeSets

A ChangeLog file is the heart of this database migration and contains all information required to change database structure and/or contents. It consists of ChangeSets that are used to define a migration. Normally ChangeSets are executed transactionally.

You have to provide a ChangeLog file that contains some ChangeSets that reflect your desired database migration. Within the sample bundle the file 'custom.changelog.xml' is referenced.

Tags

A ChangeSet can be used to tag the database at a current state. Open-Xchange uses this mechanism to tag the database to be ready for a new release.

Tags are not used in the sample custom bundle.

Preconditions

You are even able to define dependencies between ChangeSets. With that you can attach your custom ChangeSet to whatever ChangeSet you would like to. Open-Xchange uses this for ChangeSets that should be executed after the database has been tagged for a new release. All ChangeSets after tagging checks if the precondition is true. You can create complex conditions by using nested conditions. For further information have a look at the following documentation: http://www.liquibase.org/documentation/preconditions.html

Within the custom bundle the precondition is defined as follows: the changeSet with id="7.6.1:login2context:login_info", author="martin.schneider" and changeLogFile="release-7.6.1/1.login_info.changelog.xml" have to be executed (defined in ox.changelog.xml) so that all changeSets within custom.changelog.xml will become executed.

Database

Database tables

Liquibase creates two tables to handle its current state:

  • LIQUIBASECHANGELOG: Contains information about the execution status of ChangeSets. Each executed ChangeSet will be added to this table. If there is no entry available for a ChangeSet liquibase will try to execute the statement.
  • LIQUIBASECHANGELOGLOCK: Handles access to the database to not execute database migration statements twice.

Database lock

Liquibase uses a lock to only grant one liquibase instace access to the database to change. For this purpose a table named 'DATABASECHANGELOGLOCK' is created. In cluster environments the first node that successfully aquired the lock will execute the database migration statements. During this time all other nodes are blocked. After the lock was released the second node will get access to the database and read already executed ChangeSets. If all ChangeSets have been executed nothing will be done and the lock will be released. This happens with each node within the cluster.

Custom Java Classes

Within your custom ChangeLog xml file you are not only able to define changes descriptive but even execute database statements defined within a class that implements CustomSqlChange.

Have a look at the sample bundle which references and executes com.openexchange.sample.database.migration.custom.ExampleCustomSqlChange.

Command line tools

Important hints

  • changeSets are uniquely identified by its id, author and path. Use the parameter 'logicalFilePath' for every changeSet you create to set a path that can be globally identified! If you do not set a logical file path liquibase will use the physical path of the file the changeSet is located in. With that each changeSet on different cluster nodes might have been identified as a new one.
  • if you would like to use <include> statements to reference other files with liquibase statements we suggest adding parameter 'relativeToChangelogFile="true"' to avoid adding complete paths.