I'm trying to use Spring AOP to intercept methods of my GWT-RPC application (using GWT-Server library, so RPC service doesn't extend RemoteServiceServlet). When I deploy my war to tomcat and start the application, CGLIB fails for some reason. But I don't understand why CGLIB is being used for proxying at the first place. Since my RPC class implements the interface, shouldn't it be using JDK dynamic proxies?
Is there anything I need to do to debug this issue? Kindly advise.
Note: FYI, Spring encounters this exception, but I believe that's a different problem, I'm unable to understand why CGLIB proxy is in the picture.
Caused by: net.sf.cglib.core.CodeGenerationException: net.sf.ehcache.CacheException-->Another unnamed CacheManager already exists
in the same VM. Please provide unique names for each CacheManager in the config
or do one of following:
1. Use one of the CacheManager.create() static factory methods to reuse same CacheManager with same name or create one if necessary
2. Shutdown the earlier cacheManager before creating new one with same name.
Answering for the sake of other (rare) folks who might do the same mistake.
The aspect setup for spring AOP wasn't correct and was in fact trying to target almost all the classes in the context, which is why EhCache started causing problems as there were more than one CacheManager instances (because of CGLIB proxies as CacheManager doesn't implement an interface)
Related
I am trying to enable AspectJ load-time weaving (not Spring AOP) in a Spring Boot application. My goal is to weave advice into annotated fields and java.lang.reflect.Field.set(Object, Object) at load-time.
Per the Spring docs, I tried:
#Configuration
#EnableLoadTimeWeaving
public class Config {}
Running the Spring Boot application with this configuration resulted in the application context failing to load with this message:
Caused by: java.lang.IllegalStateException:
ClassLoader [jdk.internal.loader.ClassLoaders$AppClassLoader]
does NOT provide an 'addTransformer(ClassFileTransformer)' method.
Specify a custom LoadTimeWeaver or start your Java virtual machine
with Spring's agent: -javaagent:spring-instrument-{version}.jar
The latter suggestion in that message is not a good option as I am trying to avoid necessitating launch script modifications. The aspect I need to weave actually resides in a library, so all implementing Spring Boot projects will have to make whatever changes required to get LTW to work.
I also tried this configuration:
#Configuration
#EnableLoadTimeWeaving
public class Config implements LoadTimeWeavingConfigurer {
#Override
public LoadTimeWeaver getLoadTimeWeaver() {
return new ReflectiveLoadTimeWeaver();
}
}
Running the Spring Boot application with this configuration resulted in the application context failing to load with this message:
Caused by: java.lang.IllegalStateException:
ClassLoader [jdk.internal.loader.ClassLoaders$AppClassLoader]
does NOT provide an 'addTransformer(ClassFileTransformer)' method.
It seems I need to make the JVM use a class loader that has an addTransformer(ClassFileTransformer) method. I don't know how to do that, particularly for this situation. Any suggestions?
I am not an active Spring user, but I know that Spring supports annotation- or XML-configured agent hot-attachment and has some container-specific classes for that according to its documentation. It does not seem to work reliably in all situations, though, especially when running a Spring Boot application from an IDE or so.
Anyway, the AspectJ weaver 1.8.7 and more recent can be hot-attached. I explained how to do that in a Spring setup here. If you want a simpler solution with less boilerplate but one more dependency to a tiny helper library called byte-buddy-agent, you can use this solution as a shortcut. I have not tried it, but I know the helper library and am using it myself in other contexts when hot-attaching bytecode instrumentation agents, avoiding the fuss to cater to different JVM versions and configuration situations. But in order for that to work on JVM 9+, you might need to manually activate auto-attachment for the JVM, which would be another modification for your start-up script, and you would be back to square 1.
Hy,
Reading a lot about Spring AOP vs AspectJ, I still have some doubts:
1.)When using Spring AOP with classes annotated with #Aspect and using "aop:aspectj-autoproxy" tag , it can be said that we are using just the annotations of aspectj or besides that it is being used AspectJ too for the weaving?
2) Its said that AspectJ has better performance because the weaving is in compilation time, it means that the target class files are physically changed inserting the aspects in them? is it not a bit aggressive?
3)It said that Spring uses proxys for AOP, so, I undertand that when you get a bean from Spring, Spring builds a proxy in memory that has already inserted the aspects in it, right?
So why is it said that when a method from your proxy bean calls other method in the proxy, the last method will not have aspects?
Thanks
1) using aspectj-autoproxy means that #Aspectannotations are recognized, but Spring proxies are still being created, see this quote from the documentation:
Do not be misled by the name of the element:
using it will result in the creation of Spring AOP proxies. The
#AspectJ style of aspect declaration is just being used here, but the
AspectJ runtime is not involved.
2) AspectJ supports load time weaving, byte code weaving and compile time weaving. There should no difference in performance, it's just a different point in time to weave the aspect in (compilation, third party jars available, class load time), see this answer for further details.
It is actually more transparent once it's set up to have the aspects weaved at these moments, with runtime proxies there are problems when a bean calls itself using this.someMethod, the aspects don't get applied because the proxies get bypassed (#Transactional/#Secured does not work, etc.).
3) Have a look at this picture from the documentation:
With runtime proxies (non AspectJ), Spring leaves the bean class untouched. What it does is it creates a proxy that either implements the same interface as the bean (JDK proxy), or if the bean implements no interface then it dynamically creates a proxy class with CGLIB (subclass of bean).
But in both cases a proxy is created that delegates the calls to the actual bean instance. So when the bean call this.methodB() from methodA, the proxy is bypassed because the call is made directly on the bean and not on the proxy.
Spring AOP can be configured with AspectJ-sytle, ie annotations are parsed to build configuration but AspectJ compiler is not used for weaving. Only a subset of AspectJ annotations and poincut definitions can be used with Spring AOP.
Maybe, but I don't know any class that has complained. However, is possible that some classes don't allow re-weaving when modified by other bytecode tools.
Inner calls are not proxied because they call on this.method() (with this = the target bean begin proxied) and not on proxy.method() so the proxy has no chances to intercept the call. However, Spring AOP proxies usually notices when a target method return this and return itself instead, so calls like builder.withColor(Color.RED).withHat(Hat.Cowboy) will work. Note that in Spring AOP there are always two classes involved: the proxy and the target.
I want to make a call to a Spring bean (a #Component) from my message-driven bean (MDB) but have problems getting a reference to it. I've tried with a class implementing org.springframework.context.ApplicationContextAware which stores the Spring ApplicationContext in a static field in a class MyAppContext. The static field in MyAppContext is then accessed from the MDB. But MyAppContext is loaded from different classloaders. The Spring application context is correctly set in the web module classloader context, but in the MDB's classloader context, it's null.
Can I somehow instruct JBoss to use the same classloader for the web app and the MDB?
Or is there a better way than storing the Spring application context in a static field?
Thanks for any advice!
A static holder for the context is not really a good idea. To make your beans available to other applications in a Java EE environment, you should consider making use of JNDI.
Unfortunately, there is no plain JNDI exporter available out of the box, but it's fairly easy to write one yourself, as shown in this blog post: http://maestro-lab.blogspot.ro/2009/01/how-to-export-spring-managed-bean-to.html
There is however a JndiRmiServiceExporter that you may want to look at.
Once your beans are bound to names in JNDI, they can be referenced using standard CDI in your message bean without worrying about class loading issues.
Why not use "ClassPathXmlApplicationContext" to load and look up for the Spring bean you require in your MBean?
Upon starting my application, I get numerous warnings along the lines of o.s.aop.framework.Cglib2AopProxy 'Unable to proxy method [public final void org.springframework.jdbc.core.support.JdbcDaoSupport.setDataSource(javax.sql.DataSource)] because it is final: All calls to this method via a proxy will be routed directly to the proxy.' for about a dozen or so functions.
Now I perfectly understand that proxy-based aspects cannot be applied to final methods. However, I did not (on purpose, at least) try to weave any aspects into JdbcDaoSupport. I suspect it comes from <tx:annotation-driven />. Is there anything I can do to silence these warnings or, better yet, exclude those classes from the aspect weaving?
This is most likely caused by the #Transactional annotation, Spring wraps your DAO in a proxy to add the transactional behavior.
I would recommend to make your DAO implement an Interface (create and use an interface for your DAO), this will allow Spring to use a JDK dynamic proxy instead of having to use CGLib.
Using CGLIB has a limitation that methods marked as final in target class can’t be advised as final methods can’t be overridden (CGLIB creates a subclass of target class at runtime) but this limitation disappears in case of using JDK dynamic proxies.
Reference
Maybe you have extended JdbcDaoSupport and added #Transactional annotations.
You can set the Cglib2AopProxy logger to log level ERROR to avoid the warn messages. For example if using log4j and log4j.properties:
log.logger.org.springframework.aop.framework.Cglib2AopProxy = ERROR
You should use interfaces for dependency injection, the most reasons for this are described here and here.
You can read documentation about proxying mechanic for details why you see this warning.
And please vote for feature request of inspection for IntelliJ that may helps us to avoid this warnings. BTW It also contains a good explanation.
Spring Boot now uses CGLIB proxying by default, including for the AOP support. If you need interface-based proxy, you’ll need to set the spring.aop.proxy-target-class to false.
spring.aop.proxy-target-class=false
I want to use Spring AOP in my spring application. While creating AOP proxy for net.sf.ehcache.CacheManager, spring context initialization fails with the below exception:
nested exception is org.springframework.aop.framework.AopConfigException: Could not
generate CGLIB subclass of class [class net.sf.ehcache.CacheManager]: Common causes of
this problem include using a final class or a non-visible class; nested exception is
net.sf.cglib.core.CodeGenerationException: net.sf.ehcache.CacheException-->Another
unnamed CacheManager already exists in the same VM. Please provide unique names for
each CacheManager in the config or do one of following:
1. Use one of the CacheManager.create() static factory methods to reuse same
CacheManager with same name or create one if necessary
2. Shutdown the earlier cacheManager before creating new one with same name.
The source of the existing CacheManager is: InputStreamConfigurationSource
[stream=java.io.ByteArrayInputStream#955b34]
Here's my understanding about this problem - Spring is trying to create AOP proxy for net.sf.ehcache.CacheManager, and it succeeds the first time and gives a default name to the CacheManager __DEFAULT__ (found this by adding debug statements to ehcache code, building it by source and using that in my application). Now if I've multiple cache managers like 'abcCacheManager' and 'xyzCacheManager' (of type EhCacheManagerFactoryBean), Spring encounters multiple net.sf.ehcache.CacheManagers and tries to create proxy objects (something like net.sf.ehcache.CacheManager$$EnhancerByCGLIB$$b18c5958) for all of them, but with EhCache >=2.5 version, we can't have more than one caches with same name under the same VM.
I'm using EhCache 2.5.1 and would like to avoid going back to 2.4 just for this purpose. I'm not sure if this is really the problem how I can overcome this problem.
Note: Note sure if this will help, but I also noticed from the debug statements that CacheManager no-arg constructor is invoked only by the spring/CGLIB proxy generator and xyzCacheManager invokes it by passing configuration as argument.
Note : I'm answering this myself as it may help others who face the same problem.
jeha's comment on my question makes sense as I shouldn't have needed that proxy at the first place, but since I'm new to Spring AOP and proxies, I didn't know how auto-proxy mechanism works. As I modified the pointcut expressions in my advice, I didn't face the above problem after that. Prior to this, almost all the beans in the container were getting proxied and hence the problem.