Multitenancy Applications with Spring Boot and Flyway

Table Of Contents

Multitenancy applications allow different customers to work with the same application without seeing each other’s data. That means we have to set up a separate data store for each tenant. And as if that’s not hard enough, if we want to make some changes to the database, we have to do it for every tenant.

This article shows a way how to implement a Spring Boot application with a data source for each tenant and how to use Flyway to make updates to all tenant databases at once.

Example Code

This article is accompanied by a working code example on GitHub.

General Approach

To work with multiple tenants in an application we’ll have a look at:

  1. how to bind an incoming request to a tenant,
  2. how to provide the data source for the current tenant, and
  3. how to execute SQL scripts for all tenants at once.

Binding a Request to a Tenant

When the application is used by many different tenants, every tenant has their own data. This means that the business logic executed with each request sent to the application must work with the data of the tenant who sent the request.

That’s why we need to assign every request to an existing tenant.

There are different ways to bind an incoming request to a specific tenant:

  • sending a tenantId with a request as part of the URI,
  • adding a tenantId to the JWT token,
  • including a tenantId field in the header of the HTTP request,
  • and many more….

To keep it simple, let’s consider the last option. We’ll include a tenantId field in the header of the HTTP request.

In Spring Boot, to read the header from a request, we implement the WebRequestInterceptor interface. This interface allows us to intercept a request before it’s received in the web controller:

public class HeaderTenantInterceptor implements WebRequestInterceptor {

  public static final String TENANT_HEADER = "X-tenant";

  public void preHandle(WebRequest request) throws Exception {
  // other methods omitted


In the method preHandle(), we read every request’s tenantId from the header and forward it to ThreadTenantStorage.

ThreadTenantStorage is a storage that contains a ThreadLocal variable. By storing the tenantId in a ThreadLocal we can be sure that every thread has its own copy of this variable and that the current thread has no access to another tenantId:

public class ThreadTenantStorage {

  private static ThreadLocal<String> currentTenant = new ThreadLocal<>();

  public static void setTenantId(String tenantId) {

  public static String getTenantId() {
    return currentTenant.get();

  public static void clear(){

The last step in configuring the tenant binding is to make our interceptor known to Spring:

public class WebConfiguration implements WebMvcConfigurer {

  private final HeaderTenantInterceptor headerTenantInterceptor;

  public WebConfiguration(HeaderTenantInterceptor headerTenantInterceptor) {
    this.headerTenantInterceptor = headerTenantInterceptor;

  public void addInterceptors(InterceptorRegistry registry) {

Don't Use Sequential Numbers as Tenant IDs!

Sequential numbers are easy to guess. All you have to do as a client is to add or subtract from your own tenantId, modify the HTTP header, and voilá, you'll have access to another tenant's data.

Better use a UUID, as it's all but impossible to guess and people won't accidentally confuse one tenant ID with another. Better yet, verify that the logged-in user actually belongs to the specified tenant in each request.

Configuring a DataSource For Each Tenant

There are different possibilities to separate data for different tenants. We can

  • use a different schema for each tenant, or
  • use a completely different database for each tenant.

From the application’s perspective, schemas and databases are abstracted by a DataSource, so, in the code, we can handle both approaches in the same way.

In a Spring Boot application, we usually configure the DataSource in application.yaml using properties with the prefix spring.datasource. But we can define only one DataSource with these properties. To define multiple DataSources we need to use custom properties in application.yaml:

      jdbcUrl: jdbc:h2:mem:vw
      driverClassName: org.h2.Driver
      username: sa
      password: password
      jdbcUrl: jdbc:h2:mem:bmw
      driverClassName: org.h2.Driver
      username: sa
      password: password

In this case, we configured data sources for two tenants: vw and bmw.

To get access to these DataSources in our code, we can bind the properties to a Spring bean using @ConfigurationProperties:

@ConfigurationProperties(prefix = "tenants")
public class DataSourceProperties {

  private Map<Object, Object> datasources = new LinkedHashMap<>();

  public Map<Object, Object> getDatasources() {
    return datasources;

  public void setDatasources(Map<String, Map<String, String>> datasources) {
        .forEach((key, value) -> this.datasources.put(key, convert(value)));

  public DataSource convert(Map<String, String> source) {
    return DataSourceBuilder.create()

In DataSourceProperties, we build a Map with the data source names as keys and the DataSource objects as values. Now we can add a new tenant to application.yaml and the DataSource for this new tenant will be loaded automatically when the application is started.

The default configuration of Spring Boot has only one DataSource. In our case, however, we need a way to load the right data source for a tenant, depending on the tenantId from the HTTP request. We can achieve this by using an AbstractRoutingDataSource.

AbstractRoutingDataSource can manage multiple DataSources and routes between them. We can extend AbstractRoutingDataSource to route between our tenants' Datasources:

public class TenantRoutingDataSource extends AbstractRoutingDataSource {

  protected Object determineCurrentLookupKey() {
    return ThreadTenantStorage.getTenantId();


The AbstractRoutingDataSource will call determineCurrentLookupKey() whenever a client requests a connection. The current tenant is available from ThreadTenantStorage, so the method determineCurrentLookupKey() returns this current tenant. This way, TenantRoutingDataSource will find the DataSource of this tenant and return a connection to this data source automatically.

Now, we have to replace Spring Boot’s default DataSource with our TenantRoutingDataSource:

public class DataSourceConfiguration {

  private final DataSourceProperties dataSourceProperties;

  public DataSourceConfiguration(DataSourceProperties dataSourceProperties) {
    this.dataSourceProperties = dataSourceProperties;

  public DataSource dataSource() {
    TenantRoutingDataSource customDataSource = new TenantRoutingDataSource();
    return customDataSource;

To let the TenantRoutingDataSource know which DataSources to use, we pass the map DataSources from our DataSourceProperties into setTargetDataSources().

That’s it. Each HTTP request will now have its own DataSource depending on the tenantId in the HTTP header.

Migrating Multiple SQL Schemas at Once

If we want to have version control over the database state with Flyway and make changes to it like adding a column, adding a table, or dropping a constraint, we have to write SQL scripts. With Spring Boot’s Flyway support we just need to deploy the application and new scripts are executed automatically to migrate the database to the new state.

To enable Flyway for all of our tenants' data sources, first we have do disable the preconfigured property for automated Flyway migration in application.yaml:

    enabled: false

If we don’t do this, Flyway will try to migrate scripts to the current DataSource when starting the application. But during startup, we don’t have a current tenant, so ThreadTenantStorage.getTenantId() would return null and the application would crash.

Next, we want to apply the Flyway-managed SQL scripts to all DataSources we defined in the application. We can iterate over our DataSources in a @PostConstruct method:

public class DataSourceConfiguration {

  private final DataSourceProperties dataSourceProperties;

  public DataSourceConfiguration(DataSourceProperties dataSourceProperties) {
    this.dataSourceProperties = dataSourceProperties;

  public void migrate() {
    for (Object dataSource : dataSourceProperties
          .values()) {
      DataSource source = (DataSource) dataSource;
      Flyway flyway = Flyway.configure().dataSource(source).load();


Whenever the application starts, the SQL scripts are now executed for each tenant’s DataSource.

If we want to add a new tenant, we just put a new configuration in application.yaml and restart the application to trigger the SQL migration. The new tenant’s database will be updated to the current state automatically.

If we don’t want to re-compile the application for adding or removing a tenant, we can externalize the configuration of tenants (i.e. not bake application.yaml into the JAR or WAR file). Then, all it needs to trigger the Flyway migration is a restart.


Spring Boot provides good means to implement a multi-tenant application. With interceptors, it’s possible to bind the request to a tenant. Spring Boot supports working with many data sources and with Flyway we can execute SQL scripts across all of those data sources.

You can find the code examples on GitHub.

Written By:

Artur Kuksin

Written By:

Artur Kuksin

With many years of experience in software development I am always looking to learn new things. I like coding and exchanging knowledge.

Recent Posts

Guide to JUnit 5 Functional Interfaces

In this article, we will get familiar with JUnit 5 functional interfaces. JUnit 5 significantly advanced from its predecessors. Features like functional interfaces can greatly simplify our work once we grasp their functionality.

Read more

Getting Started with Spring Security and JWT

Spring Security provides a comprehensive set of security features for Java applications, covering authentication, authorization, session management, and protection against common security threats such as CSRF (Cross-Site Request Forgery).

Read more

Creating and Publishing an NPM Package with Automated Versioning and Deployment

In this step-by-step guide, we’ll create, publish, and manage an NPM package using TypeScript for better code readability and scalability. We’ll write test cases with Jest and automate our NPM package versioning and publishing process using Changesets and GitHub Actions.

Read more