That blocking behavior is what I expected but not exactly/always what I'm
getting.
When I call *(A) refresh* from another class using @Schedule (in my ejb)
the value of *count* in *refresh* is correct and statement *(B.1) *executes
before
* (A.1).*
However, when I call *(A) refresh* from an @ApplicationScoped bean (in my
war) the value of *count* in *refresh* is *incorrect* and statement
*(A.1) *executes
before* (B.1)*. Most of the time *(B.1)* doesn't execute at all. In this
scenario count always has a different count and NO exception is thrown. Its
like it just gives up.
*Method that kicks off the process *
*(A) *public void *refresh*(String usr){
// Bunch house cleaning
count = count + processRelatedProjectData();
//...
logger.log(Level.INFO, "\tExit refresh"); <-- Last statement before exit
*(A.1)*
}
*Method that uses the @Asynchronous methods*
*(B) *private Integer *processRelatedProjectData*() {
Future futureA = processProject.processData(Constant.COMPANY_A);
Future futureB = processProject.processData(Constant.COMPANY_B);
Future futureC = processProject.processData(Constant.COMPANY_C);
//... A bunch of these
try {
count = count + (Integer) futureEZLNE.get();
count = count + (Integer) futureMONKL.get();
count = count + (Integer) futureALAMN.get();
//... A bunch of these
} catch (InterruptedException | ExecutionException |
CancellationException ex) {
throw new IllegalStateException("Cannot get the answer", ex);
}
logger.log(Level.INFO, "\tEnter processRelatedProjectData"); <-- Last
statement before exit *(B.1)*
}
Dennis
On Fri, Aug 8, 2014 at 6:05 AM, Reza Rahman <Reza.Rahman_at_oracle.com> wrote:
> Please share code snippets annotating/explaining the issue. It's really
> hard to follow the description. Future.get will return as soon as there is
> a value to be returned from the underlying @Asynchronous method and should
> block until then, but no longer.
>
>
> On 8/6/2014 2:53 PM, Dennis Gesker wrote:
>
> I have an EJB method (in an EJB module within my EAR) which calls
> several other @Asynchronous methods in other classes which of course return
> a Future.
>
> In the calling method I issue get() statements so that I can use the
> returned value in a calculation. Nothing fancy there.
>
> I've been running this routine from @Schedule (once per day) method with
> no issues. Runs quickly (about 20 min) and reliably and returns the correct
> result -- an integer of record counts.
>
> However, when I try to run this same routine from my web module (within
> the same EAR) from within an @ApplicationScoped bean the method that calls
> the @Asynchronous methods returns/completes *before* the future.get()
> method called completes and returns a value. No exception is thrown.
>
> Shouldn't a Future.get() call just block/wait until its work is complete
> regardless of how its called? Is a different behavior to be expected when
> called from a web module? Perhaps some kind of implied timeout?
>
> I could use a hint.
>
> NetBeans 8 and JDK 1.8.0_11 and Glassfish 4.0.1 promoted build 10
> (July-29) on Win8.1 Pro 64
>
> Dennis
>
>
>
--
*Dennis Gesker*
<http://goog_777861789>[image: www.linkedin.com/in/gesker]
<http://www.linkedin.com/in/gesker>[image: http://www.facebook.com/gesker]
<https://www.facebook.com/gesker>[image: https://twitter.com/gesker]
<https://twitter.com/gesker>[image: http://gesker.wordpress.com]
<http://gesker.wordpress.com>[image:
https://pgp.mit.edu/pks/lookup?search=dennis%40gesker.com&op=index]
<https://pgp.mit.edu/pks/lookup?search=dennis%40gesker.com&op=index>[image:
http://www.gesker.com] <http://www.gesker.com>[image: dennis@gesker.com]
<dennis_at_gesker.com>