Jboss EAP Remote EJB invoke transaction

[erwin.etchart]

Hi all , in my company we are using Wildfly 8 , 10 and we are migrating to EAP7 using technologies EJB, JPA, WS, Activemq etc.

In this moment we are developing a couple of apps, but i have a particular problem/question. I have application A (war packaged) and application B (war packaged), both are deployed in the same sever.

App-A has a process (all EJB) that run  very fast without major problems, this process in certain moment needs to call a remote EJB (@Async) in the App-B doing a jndi lookup (because app-b can be deployed or not).

While the process of App-A runs without the ejb remote call to App-B the disk usage , cpu etc looks very good , but once the call to App-B happens the disk becomes in a very high load.

Inspecting (doing fatrace command) why this is happening i found that the directory  "standalone/data/tx-object-store/ShadowNoFileLockStore/defaultStore/StateManager/BasicAction/TwoPhaseCoordinator/AtomicAction" creates a lot of files per seconds (creating and deleting), i assume that they are files created because new transactions.

I been trying everything to disable transaction in order to avoid the file creation in this directory.

Do you know how this can be solved?

Thanks

Erwin

[wdfink]

Hello,

I don't get the problem. In general it depends on the transaction annotation whether a new transaction is created or not.

But I have never seen that the Tx manager creates a lot of files.

Could you expalin what you are doing? Maybe with the example code/project.

Did you see this in a high concurrent environment or is it just one invocation for test?

[erwin.etchart]

Thanks for your response.

App B has:

@Stateless(name="TEST")
@Remote(ITestService.class)
public class TestService implements ITestService {
  @Override
  @Asynchronous
  public Future<TestDTO> doAsync(TestDTO dto){
      return new AsyncResult<TestDTO>(execute(dto));
  }


  public TestDTO execute(TestDTO dto) {
        dto.setResult("OK")
    return dto;
  }
}

@Remote
public interface ITestService extends Serializable{
      public Future<TestDTO> doAsync(TestDTO dto);
}

App A has:

private SomeObjectResult doProcess(OtherRequest req)  {

     ITestService service = getTestService("test", "beanName");
     Future<TestDTO> future =  null;
     future = service.doAsync(new TestDTO());
     TestDTO result;
     try {
          result = future.get(20,TimeUnit.SECONDS);
     } catch (InterruptedException | ExecutionException | TimeoutException e) {
     }
return result;
}

private ITestService getTestService(String moduleName, String beanName) {
  try{
     Hashtable<String, String> jndiProperties = new Hashtable<String, String>();
       jndiProperties.put(Context.URL_PKG_PREFIXES, URL_PKG_PREFIXES);

     Context context = new InitialContext(jndiProperties);

     String appName = "";
     String distinctName = "";

     // remote view fully qualified class name
     String viewClassName = ITestService.class.getCanonicalName();
     StringBuilder lookupValue = new StringBuilder();
     lookupValue.append("ejb:");
     lookupValue.append(appName);
     lookupValue.append("/");
     lookupValue.append(moduleName);
     lookupValue.append("/");
     lookupValue.append(distinctName);
     lookupValue.append("/");
     lookupValue.append(beanName);
     lookupValue.append("!");
     lookupValue.append(viewClassName);
     return (ITestService) context.lookup(lookupValue.toString());
 }catch(Exception e){
 }
}

The doProcess method is invoked when a web service is called.

Now happens in a test environment, the files created in this object store don't get accumulated, but I'm worried for the high disk usage.

Thanks