Define different Feign client implementations based on environment - spring-boot

I have a Spring boot application which uses Feign to call an external web service via Eureka. I'd like to be able to run the application using a mocked out implementation of the Feign interface, so I can run the application locally without necessarily having Eureka or the external web service running. I had imagined defining a run configuration that allowed me to do this, but am struggling to get this working. The issue is that the Spring "magic" is defining a bean for the Feign interface no matter what I try.
Feign interface
#FeignClient(name = "http://foo-service")
public interface FooResource {
#RequestMapping(value = "/doSomething", method = GET)
String getResponse();
}
Service
public class MyService {
private FooResource fooResource;
...
public void getFoo() {
String response = this.fooResource.getResponse();
...
}
}
I tried adding a configuration class that conditionally registered a bean if the Spring profile was "local", but that was never called when I ran the application with that Spring profile:
#Configuration
public class AppConfig {
#Bean
#ConditionalOnProperty(prefix = "spring.profile", name = "active", havingValue="local")
public FooResource fooResource() {
return new FooResource() {
#Override
public String getResponse() {
return "testing";
}
};
}
}
At the point my service runs, the FooResource member variable in MyService is of type
HardCodedTarget(type=FoorResource, url=http://foo-service)
according to IntelliJ. This is the type that is automatically generated by the Spring Cloud Netflix framework, and so tries to actually communicate with the remote service.
Is there a way I can conditionally override the implementation of the Feign interface depending on a configuration setting?

the solution is like below:
public interface FeignBase {
#RequestMapping(value = "/get", method = RequestMethod.POST, headers = "Accept=application/json")
Result get(#RequestBody Token common);
}
then define your env based interface:
#Profile("prod")
#FeignClient(name = "service.name")
public interface Feign1 extends FeignBase
{}
#Profile("!prod")
#FeignClient(name = "service.name", url = "your url")
public interface Feign2 extends FeignBase
{}
finally, in your service impl:
#Resource
private FeignBase feignBase;

Having posted the same question on the Spring Cloud Netflix github repository, a useful answer was to use the Spring #Profile annotation.
I created an alternative entry point class that was not annotated with #EnabledFeignClients, and created a new configuration class that defined implementations for my Feign interfaces. This now allows me to run my application locally without the need to have Eureka running, or any dependent services.

I'm using a simpler solution to avoid having multiples interfaces for a variable parameter like url.
#FeignClient(name = "service.name", url = "${app.feign.clients.url}")
public interface YourClient{}
application-{profile}.properties
app.feign.clients.url=http://localhost:9999

Related

What is the best way to inject a singleton service into a JAX-RS/Jersey resource?

For example, what if several resource endpoints need access to some message bus to handle requests? Surely there is some way to register a singleton service class and inject it into the resources when the service class itself is NOT a resource but used by the resources.
All of the examples I've seen with providers or custom HK2 bindings refer to resources.
The closest thing I found to what I'm looking for was with this question:
Trouble creating a simple singleton class in Jersey 2 using built-in Jersey dependency injection
What is the best JAX-RS/Jersey way of doing this?
Note that the programmatic way would be most useful, I'm not using an xml file to configure the server.
If your platform supports EJB, you could use the #Singleton EJB (javax.ejb package, not javax.inject), and inject it on your resources with the #EJB annotation. Singleton EJB have also outofthebox concurrency access control.
On plain Jersey, you can use CDI application context. Declare the service class with an #ApplicationScoped annotation and inject it on your resources with #Inject. CDI will only instantiate one bean.
If you cannot annotate the service class, you can create a method that provides your service implementation an annotate it with #Produces and #ApplicationScoped.
#Produces
#ApplicationScoped
public MyService produceService() {
// instantiate your service client
}
And then use it on your resources, with:
#Inject
private MyService
Answer credit goes to #areus the answer provided here.
However, I'm providing my own answer so that I can share the code.
The Service Bean
#Singleton
public final class MyServiceBean
{
private static final AtomicInteger INSTANCES = new AtomicInteger();
private final AtomicInteger calls = new AtomicInteger();
public MyServiceBean()
{
INSTANCES.incrementAndGet();
}
public String getMessage()
{
return String.format("MyServiceBean{INSTANCES=%d, CALLED=%d}", INSTANCES.get(), calls.incrementAndGet());
}
}
The Resource Class
#Path("/messages")
public final class MyResource
{
#Inject
private MyServiceBean bean;
#GET
#Produces(MediaType.TEXT_PLAIN)
public Response handle()
{
return Response.ok(this.bean.getMessage())
.type(MediaType.TEXT_PLAIN_TYPE)
.build();
}
}
HK2 Binder
public final class MyServiceBeanBinder extends AbstractBinder
{
#Override
protected void configure()
{
bind(MyServiceBean.class).to(MyServiceBean.class).in(Singleton.class);
}
}
Then just register the binder and the resource like so:
final ResourceConfig config = new ResourceConfig();
config.register(MyResource.class);
config.register(new MyServiceBeanBinder());
Starting the server and hitting the resource multiple times yields:
MyServiceBean{INSTANCES=1, CALLED=1}
MyServiceBean{INSTANCES=1, CALLED=2}
MyServiceBean{INSTANCES=1, CALLED=3}
MyServiceBean{INSTANCES=1, CALLED=4}
MyServiceBean{INSTANCES=1, CALLED=5}

JAX-RS, Spring & ServletConfig: How to access Servlet Config in Configurator

I have troubles getting a javax.servlet.ServletConfig into a class annotated with org.springframework.context.annotation.Configuration.
My team decided that we should use spring for dependency injection and I'm trying to use it to migrate one of our simple Rest services.
My constraints are:
JAX-RS: We have several REST Services implemented JAX-RS and we don't really want to change that.
Not bound to a specific implementation of JAX-RS (Jersey & RESTEasy work fine for us and we can change from one to the other without changing underlying code)
Import as few dependencies as possible from spring: at the moment I import only org.springframework:spring-context from the spring project.
No API breakage: Deprecated is fine but the service should keep working during the transition, using our old way of doing things.
A string parameter is defined in the service's web.xml. I need to get it, instantiate a Bean with it and inject the resulting bean at several place in the code.
I don't want to mess with Spring Boot/MVC/... as the service already works and I just want the Dependency Injection part.
What I already have:
The code use javax.ws.rs.core.Application, with a class that look like that:
public class MyApplication extends Application {
#Context
private ServletConfig cfg;
public DSApplication() {
}
#Override
public Set<Class<?>> getClasses() {
return new HashSet<>();
}
#Override
public Set<Object> getSingletons() {
Set<Object> set = new HashSet<>();
String injectionStr = cfg.getInitParameter("injection");
boolean injection = false;
if (null != injectionStr && !injectionStr.isEmpty()) {
injection = Boolean.valueOf(injectionStr);
}
if (injection) {
AnnotationConfigApplicationContext context = new AnnotationConfigApplicationContext(
DSServiceProducer.class,
CContextBeanProvider.class
);
IDSService service = context.getBean(IDSService.class);
set.add(service);
} else {
set.add(new DSService()); //Old way
}
return set;
}
}
I need the servlet config in CContextBeanProvider, which look like:
#Configuration
public class CContextBeanProvider {
private ServletConfig cfg; // How to get this here ?
#Bean
public CContextBean cContextBean() {
String bean = cfg.getInitParameter("cpuContext");
return new CContextBean(bean);
}
}
CContextBean is a setting bean initialized from a string found in the web.xml of the service.
Is it possible ?
Do you have any idea how ?
Would it be easier with CDI, knowing that we run on base Tomcat ? (I've already find this if I need to use tomcat with CDI)
Could you please try to add all jersey CDI related jars to your applications ?

How to define global static header on Spring Boot Feign Client

I have a spring boot app and want to create a Feign client which has a statically defined header value (for auth, but not basic auth). I found the #Headers annotation but it doesn't seem to work in the realm of Spring Boot. My suspicion is this has something to do with it using the SpringMvcContract.
Here's the code I want to work:
#FeignClient(name = "foo", url = "http://localhost:4444/feign")
#Headers({"myHeader:value"})
public interface LocalhostClient {
But it does not add the headers.
I made a clean spring boot app with my attempts and posted to github here: github example
The only way I was able to make it work was to define the RequestInterceptor as a global bean, but I don't want to do that because it would impact other clients.
You can also achieve this by adding header to individual methods as follows:
#RequestMapping(method = RequestMethod.GET, path = "/resource", headers = {"myHeader=value"})
Using #Headers with dynamic values in Feign client + Spring Cloud (Brixton RC2) discusses a solution for dynamic values using #RequestHeader.
You can set a specific configuration class on your feign interface and define a RequestInterceptor bean in there. For example:
#FeignClient(name = "foo", url = "http://localhost:4444/feign",
configuration = FeignConfiguration.class)
public interface LocalhostClient {
}
#Configuration
public class FeignConfiguration {
#Bean
public RequestInterceptor requestTokenBearerInterceptor() {
return new RequestInterceptor() {
#Override
public void apply(RequestTemplate requestTemplate) {
// Do what you want to do
}
};
}
}
You could specify that through the application.yml file:
feign:
client:
config:
default:
defaultRequestHeaders:
Authorization:
- Basic 3ncond2dS3cr2t
otherHeader:
- value
Note that this will be applicable to all your Feign Clients if it happened that you're using more than one. If that's the case, you could add a section per client instead of adding this to the default section.
Try this
#Component
public class AuthFeignInterceptor implements RequestInterceptor {
#Override
public void apply(RequestTemplate template) {
final RequestAttributes requestAttributes = RequestContextHolder.getRequestAttributes();
if (requestAttributes != null) {
final HttpServletRequest httpServletRequest = ((ServletRequestAttributes) requestAttributes).getRequest();
template.header("Header_name","Value");
}
}
}

Simple Reverse Proxy with Spring Boot and Netflix Zuul

I'm looking to implement a simple reverse proxy with Spring Boot that is:
Easy to add routes
Ability to add custom authentication on a per route basis
Add additional headers as needed
I've looked at the facilities provided by the #EnableZuulProxy annotation but it seems too heavyweight as I don't have a desire to use Eureka, Ribbon, or Hystrix. However, #EnableZuulServer is a bit light on configuration.
Would anyone be able to provide an example of what I'm after? Is Netflix Zuul the right choice for this or is there another library I should be looking at?
Thanks!
Simple Reverse Proxy Server
It's easy to set up a simple proxy reverse using Spring Boot without Ribbon, Eureka, or Hystrix.
Simply annotate your main application class with #EnableZuulProxy and set the following property in your configuration:
ribbon.eureka.enabled=false
Then define your routes in your configuration like such:
zuul.routes.<route_name>.path=<route_path>
zuul.routes.<route_name>.url=http://<url_to_host>/
where <route_name> is an arbitrary name for your route and <route_path> is a path using Ant-style path matching.
So a concrete example would be something like this
zuul.routes.userservice.path=users/**
zuul.routes.userservice.url=http://localhost:9999/
Custom Filters
You can also implement your custom authentication and any additional headers by extending and implementing the ZuulFilter class and adding it as an #Bean to your #Configuration class.
So another concrete example:
public class MyFilter extends ZuulFilter {
#Override
public String filterType() {
// can be pre, route, post, and error
return "pre";
}
#Override
public int filterOrder() {
return 0;
}
#Override
public boolean shouldFilter() {
return true;
}
#Override
public Object run() {
// RequestContext is shared by all ZuulFilters
RequestContext ctx = RequestContext.getCurrentContext();
HttpServletRequest request = ctx.getRequest();
// add custom headers
ctx.addZuulRequestHeader("x-custom-header", "foobar");
// additional custom logic goes here
// return isn't used in current impl, null is fine
return null;
}
}
and then
#Configuration
public class GatewayApplication {
#Bean
public MyFilter myFilter() {
return new myFilter();
}
}
Zuul is a good choice. Am not sure about other alternatives but, we've started building Zuul filters (Pre/Post and Route) that could intercept the request and do all pre/post processing and route based upon your need. It is not mandatory to use the whole bunch of Eureka, Ribbon and Hysterix along with Zuul.

#Endpoint and #Transactional on the same class using Spring-ws library

I am trying to implement a web-service endpoint which would be transactional because I don't want to create a special "worker" class with transactional methods. I'm using Spring-ws library together with Spring framework.
Here is my class definition:
#Endpoint
#Transactional
#Scope(proxyMode=ScopedProxyMode.TARGET_CLASS)
public class MyEndpoint implements ApplicationContextAware { ... }
Notice that I explicitly specified proxying method to force using CGLIB.
Also notice that my class implements some interface(s), so by default Spring uses JDK dynamic proxy unless another proxying method is specified. This kind of proxies is not appropriate in my case.
The problem begins on application deployment when PayloadRootAnnotationMethodEndpointMapping class starts working. It collects names of all Spring beans with #Endpoint annotation. My endpoint class is counted twice with names "myEndpoint" and "scopedTarget.myEndpoint". This duplication causes ApplicationContextException with message "Cannot map endpoint [...] on registration key [...]: there's already endpoint [...] mapped".
Question: how can I make my endpoint class being transactional?
You might write your own PayloadRootAnnotationMethodEndpointMapping extension and override the initApplicationContext method. There you can check for the scopedTarget. prefix to filter out unwanted beans:
public class ProxiedBeanAwareEndpointMapping extends PayloadRootAnnotationMethodEndpointMapping {
#Override
protected void initApplicationContext() throws BeansException {
initInterceptors();
String[] beanNames = getApplicationContext().getBeanNamesForType(Object.class);
for (String beanName : beanNames) {
if (!beanName.startsWith("scopedTarget.")) {
Class<?> endpointClass = getApplicationContext().getType(beanName);
if (endpointClass != null && AnnotationUtils.findAnnotation(endpointClass, getEndpointAnnotationType()) != null) {
registerMethods(beanName);
}
}
}
}
}
Or you can use the open session in view approach so you don't need to proxy your #Endpoints.

Resources