Intro to Spring Remoting with HTTP Invokers – 使用HTTP Invokers的Spring Remoting介绍

最后修改: 2017年 1月 24日


1. Overview


In some cases, we need to decompose a system into several processes, each taking responsibility for a different aspect of our application. In these scenarios is not uncommon that one of the processes needs to synchronously get data from another one.


The Spring Framework offers a range of tools comprehensively called Spring Remoting that allows us to invoke remote services as if they were, at least to some extent, available locally.

Spring框架提供了一系列被全面称为Spring Remoting的工具,允许我们调用远程服务,就好像它们至少在某种程度上是本地可用的一样。

In this article, we will set up an application based on Spring’s HTTP invoker, which leverages native Java serialization and HTTP to provide remote method invocation between a client and a server application.

在这篇文章中,我们将建立一个基于Spring的 HTTP invoker的应用程序,它利用本地Java序列化和HTTP来提供客户端和服务器应用程序之间的远程方法调用。

2. Service Definition


Let’s suppose we have to implement a system that allows users to book a ride in a cab.


Let’s also suppose that we choose to build two distinct applications to obtain this goal:


  • a booking engine application to check whether a cab request can be served, and
  • a front-end web application that allows customers to book their rides, ensuring the availability of a cab has been confirmed

2.1. Service Interface


When we use Spring Remoting with HTTP invoker, we have to define our remotely callable service trough an interface to let Spring create proxies at both client and server side that encapsulate the technicalities of the remote call. So let’s start with the interface of a service that allows us to book a cab:

当我们使用Spring RemotingHTTP invoker时,我们必须通过一个接口来定义我们的可远程调用的服务,让Spring在客户端和服务器端创建代理,以封装远程调用的技术特性。因此,让我们从允许我们预订出租车的服务接口开始。

public interface CabBookingService {
    Booking bookRide(String pickUpLocation) throws BookingException;

When the service is able to allocate a cab, it returns a Booking object with a reservation code. Booking has to be a serializable because Spring’s HTTP invoker has to transfer its instances from the server to the client:


public class Booking implements Serializable {
    private String bookingCode;

    @Override public String toString() {
        return format("Ride confirmed: code '%s'.", bookingCode);

    // standard getters/setters and a constructor

If the service is not able to book a cab, a BookingException is thrown. In this case, there’s no need to mark the class as Serializable because Exception already implements it:


public class BookingException extends Exception {
    public BookingException(String message) {

2.2. Packaging the Service


The service interface along with all custom classes used as arguments, return types and exceptions have to be available in both client’s and server’s classpath. One of the most effective ways to do that is to pack all of them in a .jar file that can be later included as a dependency in the server’s and client’s pom.xml.


Let’s thus put all the code in a dedicated Maven module, called “api”; we’ll use the following Maven coordinates for this example:

因此,让我们把所有代码放在一个专门的Maven模块中,称为 “api”;本例中我们将使用以下Maven坐标。


3. Server Application


Let’s build the booking engine application to expose the service using Spring Boot.

让我们使用Spring Boot构建预订引擎应用程序来暴露服务。

3.1. Maven Dependencies


First, you’ll need to make sure your project is using Spring Boot:

首先,你需要确保你的项目使用的是Spring Boot。


You can find the last Spring Boot version here. We then need the Web starter module:

您可以在这里找到最新的Spring Boot版本。然后我们需要Web starter模块。


And we need the service definition module that we assembled in the previous step:



3.2. Service Implementation


We firstly define a class that implements the service’s interface:


public class CabBookingServiceImpl implements CabBookingService {

    @Override public Booking bookPickUp(String pickUpLocation) throws BookingException {
        if (random() < 0.3) throw new BookingException("Cab unavailable");
        return new Booking(randomUUID().toString());

Let’s pretend that this is a likely implementation. Using a test with a random value we’ll be able to reproduce both successful scenarios – when an available cab has been found and a reservation code returned – and failing scenarios – when a BookingException is thrown to indicate that there is not any available cab.


3.3. Exposing the Service


We then need to define an application with a bean of type HttpInvokerServiceExporter in the context. It will take care of exposing an HTTP entry point in the web application that will be later invoked by the client:


public class Server {

    @Bean(name = "/booking") HttpInvokerServiceExporter accountService() {
        HttpInvokerServiceExporter exporter = new HttpInvokerServiceExporter();
        exporter.setService( new CabBookingServiceImpl() );
        exporter.setServiceInterface( CabBookingService.class );
        return exporter;

    public static void main(String[] args) {, args);

It is worth noting that Spring’s HTTP invoker uses the name of the HttpInvokerServiceExporter bean as a relative path for the HTTP endpoint URL.


We can now start the server application and keep it running while we set up the client application.


4. Client Application


Let’s now write the client application.


4.1. Maven Dependencies


We’ll use the same service definition and the same Spring Boot version we used at server-side. We still need the web starter dependency, but since we don’t need to automatically start an embedded container, we can exclude the Tomcat starter from the dependency:

我们将使用相同的服务定义和我们在服务器端使用的Spring Boot版本。我们仍然需要Web启动器的依赖,但由于我们不需要自动启动一个嵌入式容器,我们可以从依赖中排除Tomcat启动器。


4.2. Client Implementation


Let’s implement the client:


public class Client {

    public HttpInvokerProxyFactoryBean invoker() {
        HttpInvokerProxyFactoryBean invoker = new HttpInvokerProxyFactoryBean();
        return invoker;

    public static void main(String[] args) throws BookingException {
        CabBookingService service = SpringApplication
          .run(Client.class, args)
        out.println(service.bookRide("13 Seagate Blvd, Key Largo, FL 33037"));

The @Bean annotated invoker() method creates an instance of HttpInvokerProxyFactoryBean. We need to provide the URL that the remote server responds at through the setServiceUrl() method.


Similarly to what we did for the server, we should also provide the interface of the service we want to invoke remotely through the setServiceInterface() method.


HttpInvokerProxyFactoryBean implements Spring’s FactoryBean. A FactoryBean is defined as a bean, but the Spring IoC container will inject the object it creates, not the factory itself. You can find more details about FactoryBean in our factory bean article.

HttpInvokerProxyFactoryBean实现了Spring的FactoryBeanFactoryBean被定义为一个Bean,但是Spring IoC容器将注入它所创建的对象,而不是工厂本身。您可以在我们的factory bean文章中找到关于FactoryBean的更多细节。

The main() method bootstraps the stand alone application and obtains an instance of CabBookingService from the context. Under the hood, this object is just a proxy created by the HttpInvokerProxyFactoryBean that takes care of all technicalities involved in the execution of the remote invocation. Thanks to it we can now easily use the proxy as we would do if the service implementation had been available locally.


Let’s run the application multiple times to execute several remote calls to verify how the client behaves when a cab is available and when it is not.


5. Caveat Emptor


When we work with technologies that allow remote invocations, there are some pitfalls we should be well aware of.


5.1. Beware of Network Related Exceptions


We should always expect the unexpected when we work with an unreliable resource as the network.


Let’s suppose the client is invoking the server while it cannot be reached – either because of a network problem or because the server is down – then Spring Remoting will raise a RemoteAccessException that is a RuntimeException.

假设客户端在调用服务器的时候,由于网络问题或服务器宕机而无法到达,那么Spring Remoting将引发一个RemoteAccessException,即一个RuntimeException。

The compiler will not then force us to include the invocation in a try-catch block, but we should always consider to do it, to properly manage network problems.


5.2. Objects Are Transferred by Value, Not by Reference


Spring Remoting HTTP marshals method arguments and returned values to transmit them on the network. This means that the server acts upon a copy of the provided argument and the client acts upon a copy of the result created by the server.

Spring Remoting HTTP将方法参数和返回的值打包,在网络上传输。这意味着服务器根据所提供的参数的副本行事,而客户端则根据服务器创建的结果的副本行事。

So we cannot expect, for instance, that invoking a method on the resulting object will change the status of the same object on the server side because there is not any shared object between client and server.


5.3. Beware of Fine-Grained Interfaces


Invoking a method across network boundaries is significantly slower than invoking it on an object in the same process.


For this reason, it is usually a good practice to define services that should be remotely invoked with coarser grained interfaces that are able to complete business transactions requiring fewer interactions, even at the expense of a more cumbersome interface.


6. Conclusion


With this example, we saw how it is easy with Spring Remoting to invoke a remote process.

通过这个例子,我们看到了Spring Remoting是如何轻松地调用一个远程进程的。

The solution is slightly less open than other widespread mechanisms like REST or web services, but in scenarios where all the components are developed with Spring, it can represent a viable and far quicker alternative.


As usual, you’ll find the sources over on GitHub.