What about the ability to filtering nodes like ClusterExecutor does?

Sebastian Łaskawiec wrote:

 

That is correct. CDI doesn't support ClusterExecutor (I created ISPN-7357 to implement this).

 

However if you want to use ClusterExecutor, just use:

@Inject

EmbeddedCacheManager ecm;

...

ecm.executor().submit(myRunnable);

Not exactly. Maybe you misunderstand me. What I mean is the runnable below is not working with CDI.

private static class TestRunnable implements Runnable, Serializable {
   @Inject
  @IdCache
   Cache<String, Object> idCache;

   @Override
   public void run() {
  System.out.println("runnable");
   System.out.println(idCache);
   output(idCache);
   System.out.println(((TestData) idCache.get("ccc")).getState());
   }
}

Here's the way you mention.

@Inject
EmbeddedCacheManager ecm;

Both below will not work with CDI. The idCache will be null.

idCache.getCacheManager().executor().submit(new TestRunnable());
ecm.executor().submit(new TestRunnable());

Instead I have to use SerializableFunction with EmbeddedCacheManager to get the idCache like below.

ecm.executor().submitConsumer(localManager -> {
   output(localManager.getCache("id-cache"));
   System.out.println(((TestData) localManager.getCache("id-cache").get("ccc")).getState());
  return null;
}, null);

GitHub - SetoKaiba/DistributedExecutionCDI

Hi. Here it is.

Just run Test or org.jboss.weld.environment.se.StartMain

There's comments mentioning when will CDI work.

Question about distributed execution thread.

Could you please have a look at the thread issue for DefaultExecutorService.

The node who calls submitEverywhere will run one by one in the same thread.

Only the other nodes will run in multiple threads concurrently.

But there is no issue with ClusterExecutor.

The reproducer is in the github project as well.

I do know the difference between DistributedExecutorService and ClusterExecutor.

I thought that what I want is ClusterExecutor. But it does not support CDI yet.

And also I took a look at CacheStream before. It's more like MapReduce which is using to process data concurrently for similar data.

Maybe it's not the stuff I'm looking for.

I'm not sure whether I'm understanding correctly.

And also, I'm using Infinispan as the Data Layer of server. But you maybe notice I'm using a cache call idCache.

And the stores the object. And the object may reference other objects by id, a reference class defined with Serializable for it.

So I maybe don't know the keys I would use in the beginning.

Instead, I want to choose nodes by load balancing myself or other stuffs I can judge myself.

That's why I'm asking for the filtering feature for DistributedExecutorService.

And also DistributedExecutorService do provide methods to run a specific node. But there's no way to run on specific nodes.

And also please don't remove DistributedExecutorService. The node choosing by keys feature is still what users want.

Anyway, I still want the feature to run on specific nodes with DistributedExecutorService.

GitHub - SetoKaiba/infinispan: Infinispan is an open source data grid platform and highly scalable NoSQL cloud data stor…

Hi. Here's maybe working fix I commit.

I imported the projects in IDEA but it seems that it can't solve the dependencies.

I can't test it. I make RunnableAdapter<T> public and judge the type during injection.

Maybe you can test whether it works or not.