We are mixing spring-integration and the scheduling capabilities from spring-boot using:
#SpringBootApplication
#EnableIntegration
#IntegrationComponentScan
#EnableConfigurationProperties
#EnableScheduling
public class MyApplication {
...
}
#EnableScheduling creates a bean named "taskScheduler" which is then used by spring-integration:
public abstract class IntegrationContextUtils {
public static final String TASK_SCHEDULER_BEAN_NAME = "taskScheduler";
...
}
private void registerTaskScheduler() {
if (!this.beanFactory.containsBean(IntegrationContextUtils.TASK_SCHEDULER_BEAN_NAME)) {
...
this.registry.registerBeanDefinition(IntegrationContextUtils.TASK_SCHEDULER_BEAN_NAME, scheduler);
}
}
Problem is, the default poolSize for spring-integration is 10 (which value is needed as we encounter starvation) while the default for spring-boot is 1 (which we also need to avoid concurrency in our scheduled processes).
Questions:
Is this a normal behavior for spring-integration to share his task scheduler bean with spring-boot scheduling capabilities?
Is there a way to specify a unique task scheduler for spring-integration, whether scheduling in boot is enabled or not?
Thanks for your answers
The behviour and logic is correct. And expectations from the convention-on-configuration from Spring Boot perspective is also correct. Only what you miss that #EnableScheduling is not a Spring Boot feature, but rather Spring Framework native: https://docs.spring.io/spring-framework/docs/current/reference/html/integration.html#scheduling. Spring Boot jsut give us an extra freedom of configuration some beans on the matter. So, we just need to rely on its auto-configuration.
If auto-configuration doesn't fit your requirements, you always can provide your own configuration and override whenever it is necessary.
Looking to the #EnableScheduling, its #Scheduled hooks and appropriate TaskSchedulingAutoConfiguration in Spring Boot, it is not so easy to override whatever you want to make Spring Integration happy at the same time. So, we should go a bit opposite direction and really override a Scheduler for Spring Integration endpoints. Every single place where you use poller, you also need to configure a custom Scheduler instead of that auto-configured.
Related
Upgrading an existing system to Spring Boot with Auto config. Currently the system is all Java config. I'm confused over whether to continue the use of #Profile. Is this annotation no longer needed? I searched extensively about upgrading and found only references to non-Spring Java migration and creating new projects.
Typical #Profile usage in our configuration classes looks something like:
#Bean
#Profile("is-standalone")
public Service unsecuredService(SomeApi someApi) {
return new ...
}
I inferred from the Spring Boot examples that using one of the #Conditional annotations is recommended like this:
#Bean
#ConditionalOnProperty("unsecured.enabled")
public Service unsecuredService(SomeApi someApi) {
return new ...
}
Then in a YAML file the is-standalone Profile enables or disables all the various properties for that Profile. Is this the proper way to upgrade? To repeat a question from above differently, can the #Profile usage be left as is? This is for a fairly large project, the upgrade is non-trivial, so I would like to do this only once!
Depends where your previous #Profile annotation is coming from. If you're using Spring's #Profile, the functionality is as follows:
Annotating a class with #Profile("dev") will load the class and register it in the Spring context only when the dev profile is active
Annotating a class with #Profile("!dev") will load the class and register it in the Spring context only when the dev profile is inactive
If this sounds like what you have already, no change is needed.
I've got a Spring WebFlux application running successfully as a standalone spring boot application.
I am attempting to run the same application in a Tomcat container, and following the documentation, I've created a class that extends AbstractReactiveWebInitializer. The class requires that I implement a method getConfigClasses that would return classes normally annotated with #Configuration. If the working spring boot app started with a class called ApplicationInitializer, then the resulting implementations would look like this:
#SpringBootApplication(scanBasePackages = "my.pkg")
#EnableDiscoveryClient
#EnableCaching
public class ApplicationInitializer {
public static void main(String... args) {
SpringApplication.run(ApplicationInitializer.class, args);
}
}
and
public class ServletInitializer extends AbstractReactiveWebInitializer {
#Override
protected Class<?>[] getConfigClasses() {
return new Class[] {ApplicationInitializer.class};
}
}
When deployed, the only thing that starts is ApplicationInitializer, none of the autoconfigured Spring Boot classes (Cloud Config, DataSource, etc) ever kick off.
The documenation states this is the class I need to implement, I just expected the remainder of the spring environment to "just work".
How should I be using this class to deploy a Reactive WebFlux Spring Boot application to a Tomcat container ?
Edit:
After some additional research, I've narrowed it down to likely just Cloud Config. During bean post processing on startup, the ConfigurationPropertiesBindingPostProcessor should be enriched with additional property sources (from cloud config), but it appears to be the default Spring properties instead, with no additional sources.
The misisng properties is causing downstream beans to fail.
Spring Boot does not support WAR packaging for Spring WebFlux applications.
The documentation you're referring to is the Spring Framework doc; Spring Framework does support that use case, but without Spring Boot.
you can extend SpringBootServletInitializer, add add reactive servlet on onStartup method
I am writing a library to provide some functionality that is shared between multiple different Spring Boot applications that I work with.
I would like to do something similar to the auto-configuration that is provided by the many Spring Boot starter libraries exist. That, or some other simple declarative way to integrate my library with the ApplicationContext of the apps using it.
I have found some resources explaining how auto configuration works. I can figure out the above problem.
However, I have not been able to find any good examples of how I can test as part of my library's test suite that it suitably integrates with a Spring Boot application. Ideally, I would start up a simple Spring Boot app written in the library's test directly just for the sake of testing, add the right annotation to it, and be able to assert that the correct beans are then configured.
I have tried creating a TestApplication class that does that and writing integration tests using the SpringBootTest annotation but the TestApplication was never started before my test started.
What can I do to start up a simple app like that solely for the purpose of testing my library? My tests are written with Spock and Spock-Spring in case that changes things versus other test frameworks.
I was able to make it work with the following test class:
#SpringBootTest
#ContextConfiguration(classes = TestApplication)
class DummyIntegrationSpec extends Specification {
#Autowired
DummyService dummyService
void 'dummy service should exist'() {
expect:
dummyService.getMessage() == DummyConfiguration.MESSAGE
}
}
and this TestApplication class at src/test/groovy/com/example/project/TestApplication.groovy
#SpringBootApplication(scanBasePackages = 'com.example.project.config')
#EnableAutoConfiguration
class TestApplication extends SpringBootServletInitializer {
#Override
protected SpringApplicationBuilder configure(SpringApplicationBuilder application) {
return application.sources(TestApplication)
}
static void main(String[] args) {
SpringApplication.run(TestApplication, args)
}
}
The two key changes I had to make in order for the TestApplication to start and load the correct context when I moved my TestApplication class from src/main to src/test were:
the TestApplication class needed to be added to the ContextConfiguration annotation
the package that my library's Java config files live in needed to be added to the SpringBootApplication scanBasePackages field
The library auto-configuration does follow a similar structure to the one mentioned in the link tom provided.
Your auto-configuration should be automatically picked while your main spring application/test is starting and all beans will be registered in your context. They will be available for auto-wiring and follow your conditions and init order.
As a summary, make sure you have an auto-configuration annotated by #Configuration class with an #Import that imports your #Configuration annotated configuration classes (inside of them you define beans with methods annotated with #Bean). Also make sure you created a spring.factories file that include your auto-configuration class and that you removed the spring boot maven plugin (for the packaging to be right).
Also, make sure your auto-configuration project is NOT annotated by things like #SpringBootApplication, #EnableAutoConfiguration, #ComponentScan or other spring boot annotations that need to be only in the main spring boot projects (There should be one of them in each stack).
Please also see the article below:
Spring boot is based on a lot of pre-made auto-configuration parent projects. You should already be familiar with spring boot starter projects.
You can easily create your own starter project by doing the following easy steps:
Create some #Configuration classes to define default beans. You should use external properties as much as possible to allow customization and try to use auto-configuration helper annotations like #AutoConfigureBefore, #AutoConfigureAfter, #ConditionalOnBean, #ConditionalOnMissingBean etc. You can find more detailed information on each annotation in the official documentation Condition annotations
Place an auto-configuration file/files that aggregates all of the #Configuration classes.
Create a file named spring.factories and place it in src/main/resources/META-INF.
In spring.factories, set org.springframework.boot.autoconfigure.EnableAutoConfiguration property with comma separated values of your #Configuration classes:
org.springframework.boot.autoconfigure.EnableAutoConfiguration=
com.mycorp.libx.autoconfigure.LibXAutoConfiguration,
com.mycorp.libx.autoconfigure.LibXWebAutoConfiguration
Using this method you can create your own auto-configuration classes that will be picked by spring-boot. Spring-boot automatically scan all maven/gradle dependencies for a spring.factories file, if it finds one, it adds all #Configuration classes specified in it to its auto-configuration process.
Make sure your auto-configuration starter project does not contain spring boot maven plugin because it will package the project as an executable JAR and won't be loaded by the classpath as intended - spring boot will not be able to find your spring.factories and won't load your configuration
This question already has answers here:
Why is my Spring #Autowired field null?
(21 answers)
Closed 7 years ago.
In order to provide POJOs in my legacy web app (Tomcat 8.0.26) with the ability to send ActiveMQ messages I've taken the recommendation to introduce Camel (2.15.2) / Spring (4.2.1) into the app to purely for the purpose of managing pooled MQ connections. I'm hoping there isn't an easier way.
Doing things the Spring way I'm thinking everything would need to be based around an MVC architecture with HTTP servlet aware controllers having access to the servlet context and therefore the Spring context in order to inject beanFactory beans into classes annotated with #Controller and #Service (and in fact there must be a Spring #Controller class that enables Spring to inject the #Service class.
However, as I've stated this is legacy code that will not be using the spring web framework.
After much pain it seems that the only way I can get beanFactory beans injected into my POJOs is to go the AspectJ and Weaving route. Before I go down this road can someone tell me that this is currently the best approach (what I've read describing this solution is from 2011 Can't get Spring to inject my dependencies - Spring Newbie) ? Can you point me to documentation and a working example?
Many thanks.
1) aspectj with #Configurable
In your #Configuration annotated class/es
you can add some more annotations
#Configuration
#EnableLoadTimeWeaving(aspectjWeaving = EnableLoadTimeWeaving.AspectJWeaving.ENABLED)
#EnableSpringConfigured
#EnableAspectJAutoProxy
to enable aspectj and the #Configurable annotation,
you need to import the aspectj lib to your project and add the spring tomcat instrumentable java agent in your tomcat lib folder (give a look here, it exaplains how to configure tomcat) http://docs.spring.io/spring-framework/docs/2.5.6/api/org/springframework/instrument/classloading/tomcat/TomcatInstrumentableClassLoader.html
this is not going to help you if you are going to create your pojos using "new"
MyPojo p = new MyPojo(); // no black magic for this, you will need to satisfies the dependencies yourself) but that would be helpful for example when you load some entities through a framework like hibernate and you want to inject something into them.. #Configurable it's an option that can be evaluated in those cases, for what you describe I would rather not use it.
2) You can have some static methods that uses some static set spring-beans and use them from your pojos, something like
class Util{
private static SprintBeanWithJmsSupport x;
public static setSpringBeanToHandleJmsMessages(SprintBeanWithJmsSupport x){
Util.x = x;
}
public static sendJmsMessage(JmsMessage m){
x.sendMessage(m)
}
}
and you can go with Util.sendJmsMessage(...)
this is a bit shitty but it does the work, I don't personally like this approach
3) set your spring beans in your pojo when they need to use them (maybe behind some nice interfaces that suit your domain)
if you go with spring mvc you will likely end up having some controllers that will use some services (generally they handle security / db access and are the entry point to start the "use cases"), as everything wthin these layers is handled by spring it will be very simple to pass the spring-bean to handle jms messaging to your pojos, this seems to me quite a nice way to handle the problem
I went mostly based on memory and something may not be completely accurate, plus my english is .. what it is, so hope this can be helpful anyway.
Can anyone explain how to do achieve a basic configuration of a task using the #Scheduled annotation without any XML configuration? All the examples I can find use at least a minimal XML configuration. For example:
http://blog.springsource.com/2010/01/05/task-scheduling-simplifications-in-spring-3-0/
This uses a typical:
<context:component-scan base-package="org/springframework/samples/task/basic/annotation"/>
<task:annotation-driven/>
So I'm just using a #Configuration annotation with a bunch of #Bean annotations. They are all instantiated at startup but the one with the #Scheduled does not run. I've used that annotation successfully in the past when using XML configuration, but never with annotations only.
Just add #EnableScheduling on you WebMvcConfig class
#Configuration
#EnableWebMvc
#EnableAsync
#EnableScheduling
public class WebMvcConfig extends WebMvcConfigurerAdapter {
/** Annotations config Stuff ... **/
}
The <task:annotation-driven /> annotation ends up declaring a ScheduledAnnotationBeanPostProcessor to read the #Scheduled annotations in your code. See here: http://static.springsource.org/spring/docs/3.1.x/javadoc-api/org/springframework/scheduling/annotation/ScheduledAnnotationBeanPostProcessor.html.
That takes care of the <task:annotation-driven /> line. To get the component scanning you'll need to use AnnotationConfigApplicationContext. Not sure if/how that works with a web container though.
In Spring 3.0, you still need that little bit of XML. However, Spring 3.1 (still in beta) introduces additional annotation options to close the gap, removing any need for XML config.
See this blog entry for how it's done. Be very careful before using beta versions of Spring in production code, though - they really are unstable.
The answers so far are all helpful for earlier versions of Spring.
Here's one that's a bit more tailored to Spring 4:
Assume that you have your main Application class annotated for Component scan like this:
#ComponentScan({"com.my.class"})
And inside of that package, you have a job class that looks like this:
#Configuration
#EnableScheduling
public class MyJobClass {
#Scheduled (cron = "* * * * * *")
public void runJob() throws DocumentException {
thingsToDoOnSchedule();
}
}
Note that the method that you annotate with #Scheduled must return void and that your cron expression needs to have 6 characters (the example shown here runs every second, which makes testing what your job does easier).
You also need the class level annotations of both #Configuration and #EnableScheduling to make this work. Either by themselves seems to get ignored.
For further reading here is the Spring 4 Enable Scheduling Reference Doc.