Is there a way to pre-load EJB instances (stateless) in JBoss 4.2.0? I know there isn't any configuration option for this, but could I write my own MBean or something to do this?
-Dave
It's probably not neccessary according to this article
in $JBOSS_HOME/server/default/conf/standardjboss.xml
you find
<container-pool-conf>
<MaximumSize>100</MaximumSize>
</container-pool-conf>
for both container-confgiurations Clustered Stateless SessionBean and
Standard Stateless SessionBean
Looking in the DTD shows MinimumSize as an optional element in container configuration for experimenting you could set the MinimumSize and check whether some server code cars about.
<!--
The container-pool-conf element holds configuration data for the
instance pool.
jboss does not read directly the subtree for this element: instead,
it is passed to the instance pool instance (if it implements
org.jboss.metadata.XmlLoadable) for it to load its parameters.
The default instance pools, EntityInstancePool and
StatelessSessionInstancePool, both accept the following configuration.
Used in: container-configuration
-->
<!ELEMENT container-pool-conf ((MinimumSize?, MaximumSize?,
strictMaximumSize?, strictTimeout?) | Synchronized)>
Related
Working with Thorntail/Kubernetes, is it possible to use dynamic values on project_defaults.yml? For example:
thorntail:
ejb3:
thread-pools:
default:
max-threads: ${my.variable}
Where my.variable will be configured on the dashboard of a specific module.
The idea is to replace standalone.xml values without performing a new deploy every time I need to change the value.
That is in theory possible, because Thorntail is based on WildFly, which allows changing some configuration values without restart. Your example refers to a WildFly configuration value, so it might be possible. Those configuration values in Thorntail that don't come from WildFly don't allow for this at all.
You can use the management fraction to let Thorntail expose the WildFly management interface. Then, you can use the WildFly management client (ModelControllerClient) to issue management operations. However, I'd discourage you from doing this, as Thorntail isn't really designed for it.
(Also, Thorntail has reached its end of life. See here for more: https://thorntail.io/posts/the-end-of-an-era/)
We can do that by configuring the variable as an environment variable. Using the previous example:
thorntail:
ejb3:
thread-pools:
default:
max-threads: ${env.my_variable:default_value}
And then we configure my_variable on the dashboard of the desired module.
It is important to notice that we should define a default value, because if that thorntail property is numeric and the environment variable is not defined on the module, we will face a parsing problem since the parser will interpret the variable name as a number. This is a real example:
Kubernets module:
RCNT_MAX_THREADS is defined with 12
project_defaults.yml:
thorntail:
ejb3:
thread-pools:
default:
max-threads: ${env.RCNT_MAX_THREADS:10}
This will cause the application to use max-threads with the value 12. If we don't define RCNT_MAX_THREADS on the module, the value 10 will be used instead.
There is also a very useful option to track what values are really being used: https://docs.thorntail.io/2.5.0.Final/#the-usage-txt-file_thorntail
I have 3 auto-generated soap services generated with Axis 1.4
I would like to set a proxy property only for one of them:
AxisProperties.setProperty("https.proxyHost",<some_host>)
However it looks like these properties are shared across all services under the same class loader
I cannot use the "https.nonProxyHost" since the hosts are dynamic
Thanks in advance...
After theral investigation I came to a conclusion that this cannot be done on AXIS 1.4:
Axis 1.4 has a bug:
problem with http proxy parameters caching mechanism
Basically, Axis 1.4 turns to a class called “DefaultHTTPSTransportClientProperties“ that acts as cache: If proxy host is null it will populate the host (in the cache) as an empty string.
The bug is that after the cache was set once, it cannot change.
On later versions of AXIS it is possible to setup different proxies:
AxisProperty uses the thread class loader hierarchy to deside which properties to use
If you want a specific property you should create your own class loader
Then spawn a thread on that class loader and instantiate AxisProperties with specific proxy settings
Now run the service from that thread and you will have specific properties for that class loader
We're migrating an EJB 3.0 application to EJB 3.1 and would like to use #Stereotype to reduce some of the EJB configuration.
The annotations we plan to have are:
#Singleton
#ConcurrencyManagement(BEAN)
#PermitAll
#Interceptors or custom #InterceptorBinding annotation
#SecurityDomain("acme") JBoss / PicketLink
Of those I know that #Singleton can't be put into a #Stereotype and has to be on the EJB itself. What else can't be put into a #Stereotype?
Update
The specification [1], [2] says that
A stereotype encapsulates any combination of:
default scope, and
a set of interceptor bindings.
The examples then use Java EE 7 #Transactional which is an #InterceptorBinding which leads me to believe that none of the above annotations can be put into a stereotype.
The The Java EE 6 Tutorial states the following:
A stereotype is a kind of annotation, applied to a bean, that
incorporates other annotations. Stereotypes can be particularly useful
in large applications where you have a number of beans that perform
similar functions. A stereotype is a kind of annotation that specifies
the following:
A default scope
Zero or more interceptor bindings
Optionally, a #Named annotation, guaranteeing default EL naming
Optionally, an #Alternative annotation, specifying that all beans with this stereotype are alternatives
So as you saw yourself, the annotations you used are not in one of the mentioned groups.
My personal suggestion is, to be careful with creating and using stereotypes, since one then always have to know (or check) what it means, so for example I prefer using #Named #RequestScoped rather than #Model because saving one line of code does not make up to not see the scope at the first glance.
In the same way that JBoss can manage jdbc data sources can it manage LDAP contexts using JNDI?
Yes, this is just a Tomcat custom <Resource>, see the Tomcat documentation.
In brief, just define a and provide your own implementation of an LdapContextFactory. JBoss will call your getObjectInstance() method with a first parameter that you can cast to a Reference, traverse all its RefAddrs, and get your option names via RefAddr.getType(), and their values via RefAddr.getContent(). Then you use whatever you have passed there to create your LdapContext and return it.
I'm trying to access a data source that is defined within a web container (JBoss) from a fat client outside the container.
I've decided to look up the data source through JNDI. Actually, my persistence framework (Ibatis) does this.
When performing queries I always end up getting this error:
java.lang.IllegalAccessException: Method=public abstract java.sql.Connection java.sql.Statement.getConnection() throws java.sql.SQLException does not return Serializable
Stacktrace:
org.jboss.resource.adapter.jdbc.remote.WrapperDataSourceService.doStatementMethod(WrapperDataSourceS
ervice.java:411),
org.jboss.resource.adapter.jdbc.remote.WrapperDataSourceService.invoke(WrapperDataSourceService.java
:223),
sun.reflect.GeneratedMethodAccessor106.invoke(Unknown Source),
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25),
java.lang.reflect.Method.invoke(Method.java:585),
org.jboss.mx.interceptor.ReflectedDispatcher.invoke(ReflectedDispatcher.java:155),
org.jboss.mx.server.Invocation.dispatch(Invocation.java:94),
org.jboss.mx.server.Invocation.invoke(Invocation.java:86),
org.jboss.mx.server.AbstractMBeanInvoker.invoke(AbstractMBeanInvoker.java:264),
org.jboss.mx.server.MBeanServerImpl.invoke(MBeanServerImpl.java:659),
My Datasource:
<?xml version="1.0" encoding="UTF-8"?>
<datasources>
<local-tx-datasource>
<jndi-name>jdbc/xxxxxDS</jndi-name>
<connection-url>jdbc:oracle:thin:#xxxxxxxxx:1521:xxxxxxx</connection-url>
<use-java-context>false</use-java-context>
<driver-class>oracle.jdbc.driver.OracleDriver</driver-class>
<user-name>xxxxxxxx</user-name>
<password>xxxxxx</password>
<exception-sorter-class-name>org.jboss.resource.adapter.jdbc.vendor.OracleExceptionSorter</exception-sorter-class-name>
<min-pool-size>5</min-pool-size>
<max-pool-size>20</max-pool-size>
</local-tx-datasource>
</datasources>
Does anyone have a clue where this could come from?
Maybe someone even knows a better way how to achieve this.
Any hints are much appreciated!
Cheers,
Michael
Not sure if this is the same issue?
JBoss DataSource config
DataSource wrappers are not usable outside of the server VM
#Michael Well, java.sql.Connection is an Interface - it might technically be possible for the concrete implementation you're getting from JBoss to be Serializable - but I don't think you're really going to have any options you can use. If it was possible, it would probably be easy :)
I think #toolkit might have said the right words with useable outside the VM - the JDBC drivers will be talking to native driver code running in the underlying OS I guess, so that might explain why you can't just pass a connection over the network elsewhere.
My advice, (if you don't get any better advice!) would be to find a different approach - if you have access to locate the resource on the JBoss directory, maybe implement a proxy object that you can locate and obtain from the directory that allows you to use the connection remotely from your fat client. That's a design pattern called data transfer object I think Wikipedia entry
#toolkit:
Well, not exactly. Since I can access the data source over JNDI, it is actually visible and thus usable.
Or am I getting something totally wrong?
#Brabster:
I think you're on the right track. Isn't there a way to make the connection serializable? Maybe it's just a configuration issue...
I've read up on Ibatis now - maybe you can make your implementations of Dao etc. Serializable, post them into your directory and so retrieve them and use them in your fat client? You'd get reuse benefits out of that too.
Here's an example of something looks similar for Wicket
JBoss wraps up all DataSources with it's own ones.
That lets it play tricks with autocommit to get the specified J2EE behaviour out of a JDBC connection. They are mostly serailizable. But you needn't trust them.
I'd look carefully at it's wrappers. I've written a surrogate for JBoss's J2EE wrappers wrapper for JDBC that works with OOCJNDI to get my DAO code unit test-able standalone.
You just wrap java.sql.Driver, point OOCJNDI at your class, and run in JUnit.
The Driver wrapper can just directly create a SQL Driver and delegate to it.
Return a java.sql.Connection wrapper of your own devising on Connect.
A ConnectionWrapper can just wrap the Connection your Oracle driver gives you,
and all it does special is set Autocommit true.
Don't forget Eclipse can wrt delgates for you. Add a member you need to delegate to , then select it and right click, source -=>add delgage methods.
This is great when you get paid by the line ;-)
Bada-bing, Bada-boom, JUnit out of the box J2EE testing.
Your problem is probably amenable to the same thing, with JUnit crossed out and FatCLient written in an crayon.
My FatClient uses RMI generated with xdoclet to talk to the J2EE server, so I don't have your problem.
I think the exception indicates that the SQLConnection object you're trying to retrieve doesn't implement the Serializable interface, so it can't be passed to you the way you asked for it.
From the limited work I've done with JDNI, if you're asking for an object via JNDI it must be serializable. As far as I know, there's no way round that - if I think of a better way I'll post it up...
OK, one obvious option is to provide a serializable object local to the datasource that uses it but doesn't have the datasource as part of its serializable object graph. The fat client could then look up that object and query it instead.
Or create a (web?) service through which to access the datasource is governed - again your fat client would hit the service - this would probably be better encapsulated and more reuseable approach if those are concerns for you.