README.md

Multi-Tenancy Sample with EclipseLink

This sample application shows how to configure EclipseLink to use the Discriminator separation strategy in a multi tenant environment.

Further Readings

Desc	Link
Solutions Guide, ch.14	http://www.eclipse.org/eclipselink/documentation/2.6/solutions/multitenancy.htm#CHDBJCJA

Technologies and frameworks

• EclipseLink 2.6.1

How it works

Configuration and usage of the Discriminator Column Value Based strategy (DCVB) is straightforward. At first we have to annotate all entity classes that are tenant aware with an @Multitenant annotation. This annotation allows us to define a separation strategy. The default strategy MultitenantType.SINGLE_TABLE fits our requirements. As next we've to tell EclipseLink the name of the discriminator column to use. This can be done in several ways, we're fine with putting more EclipseLink specific annotations on our entity classes:

@TenantDiscriminatorColumn(name = "C_TENANT_ID", contextProperty = TenantHolder.TENANT_ID)

Important to notice is the contextProperty attribute of the annotation. This tells EclipseLink to take the tenant identifier information from the configured JPA properties that are used to create the EntityManagerFactory at application startup.

    @Bean
    public EntityManagerFactory entityManagerFactory(DataSource dataSource) {
        LocalContainerEntityManagerFactoryBean factory = new LocalContainerEntityManagerFactoryBean();
        ...
        factory.getJpaPropertyMap().put(TenantHolder.TENANT_ID, TenantHolder.getTenant());
        factory.afterPropertiesSet();
        return factory.getObject();
    }

You find further details of the configuration parameters in the Javadoc of org.eclipse.persistence.config.PersistenceUnitProperties.

Restrictions

At application startup. This is not what we want. We need to pass the current tenant dynamically to EclipseLink depending on the tenant identifier we get with each http request. One can argue, that we could create new EntityManagerFactorys on the fly, cache them within a map and grab the one for the current tenant. This would work with pure JPA, but as we're using Spring Data JPA, we need to pass the correct EntityManagerFactory to Spring Data at application startup in order to build the repositories properly.

@EnableJpaRepositories(basePackageClasses = TenancyEclipseLinkSampleApplication.class,
        entityManagerFactoryRef = "entityManagerFactory")

Runtime Behavior

Let's create a new CatalogEO by sending a POST request with request body {"version":"7.2"}.

INSERT INTO T_CATALOG (C_PK, C_VERSION, C_TENANT_ID) VALUES (?, ?, ?)
	bind => [151, 7.2, FIXME]

EclipseLink recognizes that our entity class is tenant aware and adds the tenant identifier to the insert statement. The same is true when getting the data from the database. EclipseLink appends the query string with the tenant identifier,

SELECT C_PK, C_TENANT_ID, C_VERSION FROM T_CATALOG WHERE (C_TENANT_ID = ?)
	bind => [FIXME]