Wildfly 10.1.0 blocks calls to Singleton EJBs in PostConstruct

[rumpelstilzchen926]

Hi,

Wildfly 8.2.0 and 10.0.0 work fine with a Singleton Bean, which has a @Postconstruct method and in this method several threads are created with the ManagedExecutorService. These  threads call a method in another Singleton EJB, which has been injected.

This call is blocked in Wildfly 10.1.0 until the Prostconstruct thread has ended.

A simple project is attached, which demonstrated the problem and which I used to verify the problem.

Is this a bug in Wildfly 10.1.0 or is this behaviour required by the spec and Wildfly 10.0.0 and 8.2.0 were wrong to allow this?

In the real live application I am doing some serious initialization work in the @PostConstruct, because I want to be sure that the application is in a well-defined state and ready to answer requests if the deployment is successful. Thus I create as many threads as cores are available and do this intialization work in parallel. I can go back to a single-threaded variant but I am also worried about this wildfly 10.1.0 behaviour.

Any recommendations?

best regards

Dietrich

[tremes]

Hi,

I am not very experienced in this area but I guess the original behaviour is correct. EJB spec 4.8.5 Singleton Session Bean Concurrency states:

Independent of the bean’s concurrency management type, the container must ensure that no concurrent

access to the singleton session bean instance occurs until after the instance has successfully completed

its initialization sequence, including any PostConstruct lifecycle callback method(s). The con-

tainer must temporarily block any singleton session bean access attempts that arrive while the singleton

session bean is still initializing.

and TestReader @PostConstruct already happened.

[rumpelstilzchen926]

Interesting. Then this would be a Wildfly 10.1.0 bug? Where could I report it?

[tremes]

Maybe it is. You can report it at https://issues.jboss.org project WildFly. Thanks

[rumpelstilzchen926]

Thank you, Tomas. Issue created as [WFCORE-1782] Wildfly 10.1.0 blocks calls to Singleton EJBs in PostConstruct - JBoss Issue Tracker

[tremes]

I think It should go to WFLY project not WFCORE. Thanks

[rumpelstilzchen926]

You are right. As I don't know how to change category, the Wildfly developer must take care of this.

[jaikiran]

I haven't looked at the application you attached, but what exactly is blocked? The call to singleton bean B from the @PostConstruct of singleton bean A? And when you say blocked, can you get the thread dumps (a couple of them in intervals of around 3-5 seconds)? Also, is singleton bean A a @Startup singleton or a regular singleton?

[rumpelstilzchen926]

Thank you jaikiran pai for changing the category to WFLY - [WFLY-7101] Wildfly 10.1.0 blocks calls to Singleton EJBs in PostConstruct - JBoss Issue Tracker

You asked

> ... what exactly is blocked? The call to singleton bean B from the @PostConstruct of singleton bean A?

This direct call works as the suplied example code shows - the call to singleton bean B is blocked

if done from a thread created with a ManagedExecutor from the @PostConstruct of singleton bean A.

> And when you say blocked, can you get the thread dumps (a couple of them in intervals of around 3-5 seconds)?

Just execute the example I supplied - the @Postconstruct thread waits 10 seconds and during this 10 seconds

the 4 created threads are blocked and they continue to run after the @PostConstruct thread has left the method.

> Also, is singleton bean A a @Startup singleton or a regular singleton?

I made it @Startup for demonstration purpose, the bug appeared first in a non-@Startup singleton bean.