I like to prototype a JEE environment with Wildfly 10.1 and Hazelcast 3.8. Until now I only have experience with ancient JBoss 4.2.3.GA.
I already found existing resource adapter implementation based on older hazelcast 3.6 under https://github.com/hazelcast/hazelcast-ra. Unfortunately I couldn't deploy it as-is on Wildfly 10.1 since IronJacamar complained about missing equals/hashCode methods (which isn't true since they are explictly overwritten in the source code. deploying a self-built snapshot of git master had the same issue).
I ended up with migrating the ra.xml configuration code to proper javax.resource.spi annotations (#Connector, #ConfigProperty, #ConnectionDefinition) and adding javax.resource.Referenceable interface implementation (don't know whether this is necessary). The step to hazelcast 3.8 was much easier - just adding missing interface methods to HazelcastConnectionImpl.
I still struggle with deployment/configuration, so here are my questions:
How should the deployment structure for an JCA adapter look like? I tested the following approaches:
All-in-one: RAR file containing all of cache-api-1.0.0.jar, hazelcast-3.8.jar, hazelcast-client-3.8.jar, my-hazelcast-ra-impl.jar and deployment descriptors.
By-Library: Added new modules javax.cache.api (cache-api-1.0.0.jar) and com.hazelcast.hazelcast (hazelcast-3.8.jar, hazelcast-client-3.8.jar) to ${WILDFLY_HOME}/modules/ and declared appropriate module dependencies in jboss-deployment-structure.xml. RAR file contains my-hazelcast-ra-impl.jar, hazelcast.xml and deployment descriptors.
By-Adapter: Added a new module my.hazelcast.ra (cache-api-1.0.0.jar, my-hazelcast-ra-impl.jar) to ${WILDFLY_HOME}/modules/ and declared appropriate module dependency in jboss-deployment-structure.xml. RAR file contains hazelcast-3.8.jar, hazelcast-client-3.8.jar, hazelcast.xml and deployment descriptors.
Where is the proper place to deploy a hazelcast.xml configuration file into Wildfly 10.1? It seems that I need to pack it next to ResourceAdapterImpl class (my-hazelcast-ra-impl.jar) so that the class loader finds it and prefers it over hazelcast-default.xml. It contains only global configuration options like group/network. No cache definitions since caches should be configured/created on-demand via CDI.
Is there something like a conf folder where I can deploy hazelcast.xml file separate from binary RAR contents? It would be nice if it could be hot-deployed (for prototyping) but that is not mandatory.
Should it be placed somehow inside subsystem configurations within standalone.xml? I found similar cache-container configurations for infinispan subsystem but don't know how to adapt this to hazelcast (since it's not an own subsystem).
In Wildfly Management webinterface I can find the deployed RAR under Depoyments and in the JNDI view, but it is not listed under Configuration -> Subsystems -> Resource Adapters. I can create a new entry there but don't find any advantage. What is the meaning of this configuration option?
Thank you in advance
Related
Reading the documentation of redhat (https://access.redhat.com/site/documentation/en-US/JBoss_Enterprise_Application_Platform/6/html/Development_Guide/chap-Class_Loading_and_Modules.html) I found that the application server classloader
has a priority list when loading classes that are used to avoid any conflict between
loaded classes, The order is as below
Implicit dependencies.
These are the dependencies that are added automatically by JBoss Enterprise Application Platform 6, such as the JAVA EE APIs. These dependencies have the highest class loader precedence because they contain common functionality and APIs that are supplied by JBoss Enterprise Application Platform 6.
Refer to Section 3.7.1, “Implicit Module Dependencies” for complete details about each implicit dependency.
Explicit dependencies.
These are dependencies that are manually added in the application configuration. This can be done using the application's MANIFEST.MF file or the new optional JBoss deployment descriptor jboss-deployment-structure.xml file.
Refer to Section 3.2, “Add an Explicit Module Dependency to a Deployment” to learn how to add explicit dependencies.
Local resources.
Class files packaged up inside the deployment itself, e.g. from the WEB-INF/classes or WEB-INF/lib directories of a WAR file.
Inter-deployment dependencies.
These are dependencies on other deployments in a EAR deployment. This can include classes in the lib directory of the EAR or classes defined in other EJB jars.
I tried to test this order by using a JSF webapp (rich faces) in my EAR archive
My ear is as below :
sample.ear
--- sport.war
--- mysql.jar
--- lib
Usescase 1 : I added the JSF jars under the webapp (sport.war/WEB-INF/lib): [jsf-api-2.1.14.jar/jsf-impl-2.1.14.jar/portletbridge-api-3.1.2.Final.jar/portletbridge-impl-3.1.2.Final.jar], the jboss server started well and I don't have any exception
Usescase2: I added the JSF jars under sample.ear/lib
==> When I start the jboss server I get an exception (it sounds that the application server loaded the module JSF provided by jboss Implicit dependencies instead of the one in my sample.ear/lib)
I can't understand why in the 1srt usescase the Class Loading Precedence is not respected while in the 2sd usescase the Class Loading Precedence is respected?
Could you please clarify me this point
ENV
JBoss EAP 6.1.0.GA (AS 7.2.0.Final-redhat-8)
JDK 6
Without seeing the exact deployment exception that you got it is difficult to diagnose the issue.
In the first scenario the packaged libraries are loaded in the same class loader as your application.
In the second scenario the packaged libraries are loaded in a separate module and class loader.
The above means that , The deployment issue you were having fo not have to be related to Class Loading Precedence they could also be related to Class Loading Isolation.
Also Jboss and EAP already come with a prepackaged implementation of JSF, and you might be experiencing collisions due to version mismatch
If you were looking to replace the default JSF implementation on JBoss the better option to do so would have been to put the new JSF implementation in a static module, just like the default one , and have Jboss load it on demand.
I'm developing a Liferay application, consisting on 2 different portlets, an both have to make certain operations in common, so I decided to put that operations in static methods in an external Utils class.
I have to externalize that class to avoid duplicating the same code in both portlets, and I want to have the portlets in different WAR files.
I know I can package the Utils class in a JAR file, but we are still developing and we don't want to regenerate the JAR and restart the Tomcat for every change.
Which is the best option and how can I perform it?
If you're using the Liferay SDK, you can use the clients (recently changed to shared) directory to put your common code.
A good example is how deploy-listener-shared is used in conjunction with deploy-listener-hook.
From what it looks like, all the configuration you need to do is to modify your build.xml files that will use the client\shared classes. If you look at build file of deploy-listener-hook you can see all you need to add is the.
For the new SDK:
<property name="import.shared" value="my-utils-shared" />
For the older SDK:
<property name="dependent.clients" value="my-utils-client" />
Hope this helps!
There is another method that involves building a JAR file but it doesn't require a server restart (on Tomcat at least).
Write a build script for your JAR file so it compiles, builds the JAR and finally copies it to the following location:
{tomcat}/webapps/ROOT/WEB-INF/lib
Then in your portlet open the "liferay-plugin-package.properties" (in Liferay Developer Studio / Liferay IDE this should open with a nice GUI).
Then add the name of your JAR to the "portal-dependency-jars" list in this file so in the source it would like (Or just hit the "Add" button in the GUI and select the JARs you want):
portal-dependency-jars=my-custom-lib.jar,my-other-custom-lib.jar
Save the file, and redeploy the portlet, and the JAR will be copied across when the portlet is deployed.
I've used this method for custom JARs, and 3rd party JARs that I've needed to use in my portlets.
For the development phase just package the jar file with both applications.
Unless one application depends on the other somehow it is completely ok.
Another solution is to use JRebel tool. It will allow you to redeploy jar in tomcat without restarting.
Also you may have several portlets in one .war. You may just define them both in portlet.xml.
I have couple of projects embedded in a web app as jars. Each project has a log4j.properties file. When the web app is deloyed, which configuration file gets used and how to override the configurations in log4j.xml in a jar file. The jars are not web projects. They are more like service layer code. What is the order in which the log4j.properties file is loaded in the below scenario
Web-project
classes
log4j.properties
ProjectB.jar
com
log4j.properties
ProjectC.jar
com
log4j.properties and so on.
If your jars are separate web applications, each web application should use the one it first finds on the classpath (WEB-INF/classes).
You can pass a -Dlog4j.configuration=path_to_file setting to e.g. the tomcat startup to make sure that it uses the one you intend it to use.
However, this would then to my understanding and knowledge be the one that tomcat will use for every webapp that is deployed.
Question here is how you deploy your apps. Either all web applications in one tomcat in which case you probably want each web application to use a different log4.properties (or log4j.xml) or in the case where you specify one to tomcat, it should use the one you specify.
What it boils down to as far as i know: Either the first one found on classpath (remember: each web-app has it's own classpath) or the one you specify via the -D setting.
Just found this reference which i think nicely summarizes the main concepts of logging in tomcat and webapps deployed in tomcat: http://wiki.apache.org/tomcat/FAQ/Logging
If you need even more control over the log4j logging, you can resort to coding the log4j configuration in java. However, this would mean that you have to modify the source code and add code into it, which relates to infrastructure and relates deployment details to your application (not so nice).
If you set additivity to false in common packages at ProjectA, ProjectB and WebProject your log will not duplicate.
log4j.additivity.[logged package] = false
For example:
log4j.properies -> Project A, Project B
log4j.additivity.org.spring.framework = false
All org.spring.framework log will come from WebProject ignoring ProjectA and ProjectB.
I want to deploy some different applications using one jboss (jboss as 7). Can anyone make it clear for me if its possible to set one log4j configuration (log4j.xml) for multiple wars and ears or i'll have to put a copy of configuration into each archive?
Or maybe someone can suggest a differeng logging engine, more native to jboss7?
log4j configuration is always global to runtime. ie. A JBoss instance can't have multiple log4j configuration. log4j can be initialized or reinitialized with a single config file. it can be a simple properties file or a XML incompliance with log4j.dtd packaged with log4j*.jar. start the jboss instance with -Dlog4j.configuration=/anypath/log4jconfig.xml
I do not believe there is a way to configure log4j globally. JBoss AS7 uses JBoss Logging for it's log provider. You can configure the logging subsystem is in the standalone.xml or the domain.xml depending on if you're running in standalone or domain mode.
The documentation is not greatest unfortunately, but if you install the jboss-as-logging_1_1.xsd in your IDE you can get auto-complete. All the schemas are located in the docs/schema under the installed directory.
I have an app that uses apache-commons-collections v3.2.1. When I deploy the war file using the WAS 6.1 Integrated Solutions Console I get errors because the commons-collections.jar file in isclite.ear gets loaded before the one that I bundle with my .war file. The problem is that my application requires v3.0.0+ of commons collections. The one in isclite.ear is version 2.1.
So, can I deploy into WAS 6.1 without involving isclite? Can I just drop a war file into a directory somewhere? Is there an option I can select in the process of deploying through the admin interface to skip or remove the isclite.ear dependency?
Thanks.
I had a very similar problem with ColdFusion where a jar in the server was a different version than the one that was needed by my application. I had to call a non-delegating loader (one that doesn't ask the parent to first load the class before loading the requested jar). The loader I used is open source. You can find out about it at
http://groups.google.com/group/javaloader-dev