Multi-Tenancy

in Java applications

@ladislavGazo

gazo@seges.sk

The application

user

bunch of JavaScript

service

domain model

Implication

user1 ... userN

same bunch of JavaScript

service

ONE database

One application to rule them all

The term multi-tenancy in general is applied to software development to indicate an architecture in which a single running instance of an application simultaneously serves multiple clients (tenants).

This is highly common in SaaS solutions.

Isolating information (data, customizations, etc) pertaining to the various tenants is a particular challenge in these systems.

Split me now!

user1 -> BA, user2 -> BA, user 3 -> KE, user 4 -> ZA

again the same JavaScript

(almost) the same service

DB (BA), DB (KE), DB (ZA)

a lot of "tenant" dependent data - high volumes
performance
separation -> easier backup and restore
scalability of front-end and service layer

Hibernate is the easiest

Multi-tenant configuration

<property name="hibernate.multiTenancy" value="DATABASE" />
<property name="hibernate.tenant_identifier_resolver" value="sk.seges.hroddelenie.configuration.MultiTenantIdentifierResolver" />

<property name="hibernate.multi_tenant_connection_provider" value="org.hibernate.service.jdbc.connections.spi.DataSourceBasedMultiTenantConnectionProviderImpl" />


<property name="hibernate.connection.datasource" value="java:comp/env/jdbc/hr" />

<property name="hibernate.multi_tenant.datasource.identifier_for_any" value="default" />

Switch it ON

hibernate.multiTenancy

Configure where to look for the current tenant identifier

hibernate.tenant_identifier_resolver

Where are my connections?

hibernate.multi_tenant_connection_provider

Hibernate's DataSourceBasedMultiTenantConnectionProviderImpl utilizes JNDI lookups

Point to JNDI root for datasources

hibernate.connection.datasource

... and the default

hibernate.multi_tenant.datasource.identifier_for_any

So in the end = java:comp/env/jdbc/hr/default

But there are things that don't work

EHCache configuration from previous version is different
Hibernate's internal schema update does not work

NPE !!! ........ who would have said that

Migration

Hibernate exporter

is not helpful

Do you have Envers?

then you need custom exporter

It is easy to write

but don't forget the AuditConfiguration

Ejb3Configuration jpaConfiguration = new Ejb3Configuration().configure(				unitName, null);
Configuration hibernateConfiguration = jpaConfiguration
				.getHibernateConfiguration();
		
hibernateConfiguration.buildMappings();
AuditConfiguration auditConfiguration = AuditConfiguration.getFor(hibernateConfiguration);

and use EnversSchemaGenerator at last

Get rid of NPE bug

 { "hibernate.multiTenancy",
   "hibernate.tenant_identifier_resolver",
   "hibernate.multi_tenant_connection_provider",
   "hibernate.connection.datasource",
   "hibernate.multi_tenant.datasource.identifier_for_any" }

Run the exporter from Maven

Combine it with Liquibase

version 3.0.6 contains feature to populate all tenant databases

                 MultiTenantSpringLiquibase

Don't forget...

it is maintenable but generated SQL needs to be checked

existing databases must be moved to "point 0"

A bit of logging

http://www.slf4j.org/faq.html#logging_performance

Importance of logging - using MDC to track requests

                           MDC.put("tenant", clientOrgz);

 log4j.appender.stdout.layout.ConversionPattern = APP [%d{ISO8601}] - %-5p [%t|%X{tenant}|%X{requestId}] %F:%M %-30.30c{1} - %m%n

And the most boring...

Design

your app to be multi-tenant

Send tenant identification

in HTTP Header

use Filter to get it out

in Service parameters
using ThreadLocal variable
via tenant-URL-based service calls

rmi://localhost:12345/<tenant>/<service>

etc...

Thanks for listening

@ladislavGazo

gazo@seges.sk