(Quick Reference)

4.4.5 Multiple Datasources - Reference Documentation

Authors: Graeme Rocher, Peter Ledbrook, Marc Palmer, Jeff Brown, Luke Daley, Burt Beckwith, Lari Hotari

Version: 3.1.4

4.4.5 Multiple Datasources

By default all domain classes share a single DataSource and a single database, but you have the option to partition your domain classes into two or more DataSources.

Configuring Additional DataSources

The default DataSource configuration in grails-app/conf/application.yml looks something like this:

---
dataSource:
    pooled: true
    jmxExport: true
    driverClassName: org.h2.Driver
    username: sa
    password:

environments: development: dataSource: dbCreate: create-drop url: jdbc:h2:mem:devDb;MVCC=TRUE;LOCK_TIMEOUT=10000;DB_CLOSE_ON_EXIT=FALSE test: dataSource: dbCreate: update url: jdbc:h2:mem:testDb;MVCC=TRUE;LOCK_TIMEOUT=10000;DB_CLOSE_ON_EXIT=FALSE production: dataSource: dbCreate: update url: jdbc:h2:prodDb;MVCC=TRUE;LOCK_TIMEOUT=10000;DB_CLOSE_ON_EXIT=FALSE properties: jmxEnabled: true initialSize: 5

This configures a single DataSource with the Spring bean named dataSource. To configure extra DataSources, add a dataSources block (at the top level, in an environment block, or both, just like the standard DataSource definition) with a custom name. For example, this configuration adds a second DataSource, using MySQL in the development environment and Oracle in production:

---
dataSources:
    dataSource:
        pooled: true
        jmxExport: true
        driverClassName: org.h2.Driver
        username: sa
        password:
    lookup:
        dialect: org.hibernate.dialect.MySQLInnoDBDialect
        driverClassName: com.mysql.jdbc.Driver
        username: lookup
        password: secret
        url: jdbc:mysql://localhost/lookup
        dbCreate: update

environments: development: dataSources: dataSource: dbCreate: create-drop url: jdbc:h2:mem:devDb;MVCC=TRUE;LOCK_TIMEOUT=10000;DB_CLOSE_ON_EXIT=FALSE test: dataSources: dataSource: dbCreate: update url: jdbc:h2:mem:testDb;MVCC=TRUE;LOCK_TIMEOUT=10000;DB_CLOSE_ON_EXIT=FALSE production: dataSources: dataSource: dbCreate: update url: jdbc:h2:prodDb;MVCC=TRUE;LOCK_TIMEOUT=10000;DB_CLOSE_ON_EXIT=FALSE properties: jmxEnabled: true initialSize: 5 … lookup: dialect: org.hibernate.dialect.Oracle10gDialect driverClassName: oracle.jdbc.driver.OracleDriver username: lookup password: secret url: jdbc:oracle:thin:@localhost:1521:lookup dbCreate: update

You can use the same or different databases as long as they're supported by Hibernate.

Configuring Domain Classes

If a domain class has no DataSource configuration, it defaults to the standard 'dataSource'. Set the datasource property in the mapping block to configure a non-default DataSource. For example, if you want to use the ZipCode domain to use the 'lookup' DataSource, configure it like this:

class ZipCode {

String code

static mapping = { datasource 'lookup' } }

A domain class can also use two or more DataSources. Use the datasources property with a list of names to configure more than one, for example:

class ZipCode {

String code

static mapping = { datasources(['lookup', 'auditing']) } }

If a domain class uses the default DataSource and one or more others, use the special name 'DEFAULT' to indicate the default DataSource:

class ZipCode {

String code

static mapping = { datasources(['lookup', 'DEFAULT']) } }

If a domain class uses all configured DataSources use the special value 'ALL':

class ZipCode {

String code

static mapping = { datasource 'ALL' } }

Namespaces and GORM Methods

If a domain class uses more than one DataSource then you can use the namespace implied by each DataSource name to make GORM calls for a particular DataSource. For example, consider this class which uses two DataSources:

class ZipCode {

String code

static mapping = { datasources(['lookup', 'auditing']) } }

The first DataSource specified is the default when not using an explicit namespace, so in this case we default to 'lookup'. But you can call GORM methods on the 'auditing' DataSource with the DataSource name, for example:

def zipCode = ZipCode.auditing.get(42)
…
zipCode.auditing.save()

As you can see, you add the DataSource to the method call in both the static case and the instance case.

Hibernate Mapped Domain Classes

You can also partition annotated Java classes into separate datasources. Classes using the default datasource are registered in grails-app/conf/hibernate.cfg.xml. To specify that an annotated class uses a non-default datasource, create a hibernate.cfg.xml file for that datasource with the file name prefixed with the datasource name.

For example if the Book class is in the default datasource, you would register that in grails-app/conf/hibernate.cfg.xml:

<?xml version='1.0' encoding='UTF-8'?>
<!DOCTYPE hibernate-configuration PUBLIC
          '-//Hibernate/Hibernate Configuration DTD 3.0//EN'
          'http://hibernate.sourceforge.net/hibernate-configuration-3.0.dtd'>
<hibernate-configuration>
   <session-factory>
      <mapping class='org.example.Book'/>
   </session-factory>
</hibernate-configuration>

and if the Library class is in the "ds2" datasource, you would register that in grails-app/conf/ds2_hibernate.cfg.xml:

<?xml version='1.0' encoding='UTF-8'?>
<!DOCTYPE hibernate-configuration PUBLIC
          '-//Hibernate/Hibernate Configuration DTD 3.0//EN'
          'http://hibernate.sourceforge.net/hibernate-configuration-3.0.dtd'>
<hibernate-configuration>
   <session-factory>
      <mapping class='org.example.Library'/>
   </session-factory>
</hibernate-configuration>

The process is the same for classes mapped with hbm.xml files - just list them in the appropriate hibernate.cfg.xml file.

Services

Like Domain classes, by default Services use the default DataSource and PlatformTransactionManager. To configure a Service to use a different DataSource, use the static datasource property, for example:

class DataService {

static datasource = 'lookup'

void someMethod(...) { … } }

A transactional service can only use a single DataSource, so be sure to only make changes for domain classes whose DataSource is the same as the Service.

Note that the datasource specified in a service has no bearing on which datasources are used for domain classes; that's determined by their declared datasources in the domain classes themselves. It's used to declare which transaction manager to use.

What you'll see is that if you have a Foo domain class in dataSource1 and a Bar domain class in dataSource2, and WahooService uses dataSource1, a service method that saves a new Foo and a new Bar will only be transactional for Foo since they share the datasource. The transaction won't affect the Bar instance. If you want both to be transactional you'd need to use two services and XA datasources for two-phase commit, e.g. with the Atomikos plugin.

Transactions across multiple datasources

Grails uses the Best Efforts 1PC pattern for handling transactions across multiple datasources.

The Best Efforts 1PC pattern is fairly general but can fail in some circumstances that the developer must be aware of. This is a non-XA pattern that involves a synchronized single-phase commit of a number of resources. Because the 2PC is not used, it can never be as safe as an XA transaction, but is often good enough if the participants are aware of the compromises.

The basic idea is to delay the commit of all resources as late as possible in a transaction so that the only thing that can go wrong is an infrastructure failure (not a business-processing error). Systems that rely on Best Efforts 1PC reason that infrastructure failures are rare enough that they can afford to take the risk in return for higher throughput. If business-processing services are also designed to be idempotent, then little can go wrong in practice.

The BE1PC implementation was added in Grails 2.3.6. . Before this change additional datasources didn't take part in transactions initiated in Grails. The transactions in additional datasources were basically in auto commit mode. In some cases this might be the wanted behavior. One reason might be performance: on the start of each new transaction, the BE1PC transaction manager creates a new transaction to each datasource. It's possible to leave an additional datasource out of the BE1PC transaction manager by setting transactional = false in the respective configuration block of the additional dataSource. Datasources with readOnly = true will also be left out of the chained transaction manager (since 2.3.7).

By default, the BE1PC implementation will add all beans implementing the Spring PlatformTransactionManager interface to the chained BE1PC transaction manager. For example, a possible JMSTransactionManager bean in the Grails application context would be added to the Grails BE1PC transaction manager's chain of transaction managers.

You can exclude transaction manager beans from the BE1PC implementation with the this configuration option:

grails.transaction.chainedTransactionManagerPostProcessor.blacklistPattern = '.*'
The exclude matching is done on the name of the transaction manager bean. The transaction managers of datasources with transactional = false or readOnly = true will be skipped and using this configuration option is not required in that case.

XA and Two-phase Commit

When the Best Efforts 1PC pattern isn't suitable for handling transactions across multiple transactional resources (not only datasources), there are several options available for adding XA/2PC support to Grails applications.

The Spring transactions documentation contains information about integrating the JTA/XA transaction manager of different application servers. In this case, you can configure a bean with the name transactionManager manually in resources.groovy or resources.xml file.

There is also Atomikos plugin available for XA support in Grails applications.