# Failing in one branch of the akka stream did not failed execution of another branch

**URL:** <https://discuss.akka.io/t/failing-in-one-branch-of-the-akka-stream-did-not-failed-execution-of-another-branch/7307>\
**Category:** Akka Streams & Alpakka\
**Tags:** streams\
**Created:** [October 8, 2020, 12:41pm UTC](https://discuss.akka.io/t/failing-in-one-branch-of-the-akka-stream-did-not-failed-execution-of-another-branch/7307 "2020-10-08T12:41:11Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![dparamoshkin](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.akka.io/dparamoshkin/32/2053_2.png) [@dparamoshkin](https://discuss.akka.io/u/dparamoshkin)\
**Post date:** [October 8, 2020, 12:41pm UTC](https://discuss.akka.io/t/failing-in-one-branch-of-the-akka-stream-did-not-failed-execution-of-another-branch/7307/1 "2020-10-08T12:41:11Z")

</div>

I have quite simple Akka Graph with one source and 2 sinks.  
What I expect if sink in one branch fails it should stop executing second branch, but it’s not what I see in reality. In reality second branch is continuing to consume rest of the tuples. Do I miss anything?

```auto
     Sink<Integer, CompletionStage<Done>> topHeadSink = Sink.foreach(o -> {
      System.out.println("Upper sink " + o);
      throw new RuntimeException("UPS");
    });
    Sink<Integer, CompletionStage<Done>> bottomHeadSink = Sink.foreach( o -> System.out.println("Bottom sink " + o));

    final RunnableGraph<Pair<CompletionStage<Done>, CompletionStage<Done>>> g =
        RunnableGraph.<Pair<CompletionStage<Done>, CompletionStage<Done>>>fromGraph(
            GraphDSL.create(
                topHeadSink, // import this sink into the graph
                bottomHeadSink, // and this as well
                Keep.both(),
                (b, top, bottom) -> {
                  final UniformFanOutShape<Integer, Integer> bcast = b.add(Broadcast.create(2));

                  b.from(b.add(Source.from(Arrays.asList(1, 2, 3, 4, 5, 6, 7, 8, 9, 10))))
                      .viaFanOut(bcast)
                      .to(top);
                  b.from(bcast).toInlet(bottom.in());
                  return ClosedShape.getInstance();
                }));

    g.run(_runner.actorSystem());

```

Result of executing this graph is

```auto
Upper sink 1
Bottom sink 1
Bottom sink 2
Bottom sink 3
Bottom sink 4
Bottom sink 5
Bottom sink 6
Bottom sink 7
Bottom sink 8
Bottom sink 9
Bottom sink 10

```

---

<div class="post-metadata">

**Author:** ![tg44](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.akka.io/tg44/32/1099_2.png) [@tg44](https://discuss.akka.io/u/tg44)\
**Post date:** [October 8, 2020, 4:35pm UTC](https://discuss.akka.io/t/failing-in-one-branch-of-the-akka-stream-did-not-failed-execution-of-another-branch/7307/2 "2020-10-08T16:35:44Z")

</div>

> **[Broadcast • Akka Documentation](https://doc.akka.io/docs/akka/current/stream/operators/Broadcast.html#reactive-streams-semantics)**
>
> Akka is a toolkit for building highly concurrent, distributed, and resilient message-driven applications for Java and Scala.

> **cancels** depends on the `eagerCancel` flag. If it is true, when any downstream cancels, if false, when all downstreams cancel.

You can create a broadcast with an `eagerCancel` flag, if you set that to true, the BC will stop if any output stops.