Is ForkJoinPool tracable?

[no_jihun]

With this like code, I couldn't see details inside callable.

code

public class ApmTest {

    static ForkJoinPool fjp = new ForkJoinPool();
    static ExecutorService pool = Executors.newFixedThreadPool(10);

    public static void main(String[] args) throws ExecutionException, InterruptedException {
        String result1 = fjp.submit(() -> runInForkJoinPool()).get();
        String result2 = pool.submit(() -> runInExecutorService()).get();
        System.out.println("result is " + result1);
        System.out.println("result is " + result2);
        fjp.shutdown();
        pool.shutdown();

    }

    public static String runInExecutorService() {
        try {
            Thread.sleep(1000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        return "runInExecutorService";
    }

    public static String runInForkJoinPool() {
        try {
            Thread.sleep(1000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        return "runInForkJoinPool";
    }
}

result

configuration

    -javaagent:path of elastic-apm-agent-1.13.0.jar \
    -Delastic.apm.service_name=myservicename \
    -Delastic.apm.server_urls=myserverurl \
    -Delastic.apm.trace_methods=mypackage.ApmTest#* \
    -Delastic.apm.transaction_max_spans=50000 \
    -Delastic.apm.trace_methods_duration_threshold=50ms \
    -Delastic.apm.profiling_spans_enabled=true \
    -Delastic.apm.profiling_included_classes=mypackage.* \
    -Delastic.apm.application_packages=mypackage \

Backgrounds

Because there are some tasks that are need to be executed in FrokJoinPool concepts; like parelllStream().map(...) at deep level operation, So I run whole task in ForkJoinPool.

Question

I hope transaction is started from main and the followed method calls those executed inside ForkJoinPool are profiled as in same transaction.

ForkJoinPool is not possible to trace?

Or, Any workaround to manually propagate context to each parallel executions?

Update

When I execute methods in stream() it profiled, but methods executed in parallelStream() are not profiled.

[Eyal_Koren]

Hi and thanks for the question.

Inter-thread context propagation is not supported for ForkJoinPool.
There are two ways you can make that happen:

1. wrap Callable/Runnable/Task submitted to the pool with your own type that also contains the Span, which you then activate on the worker thread through our public API, similar to the example demonstrating context propagation through blocking queues. If you do that with your example, feel free to make a pull request to add this so it can be a reference example to others as well
2. add a dedicated plugin, that will work similarly to our Executors plugin (likely a lot more complicated).

parallelStream is not supported as well. You may be able to support that by wrapping your collection elements as well alongside a Span, and activate/deactivate/end this Span when processing the collection element on the worker thread.

I hope this helps.

[Eyal_Koren]

@no_jihun Would you like to test our new ForkJoinPool support? If so, please try this snapshot build and let us know.