JarCasting
OpenTracing Akka Instrumentation
OpenTracing instrumentation for Akka.
Installation
pom.xml
<dependency>
<groupId>io.opentracing.contrib</groupId>
<artifactId>opentracing-akka</artifactId>
<version>VERSION</version>
</dependency>
Usage
Actor's Span propagation (experimental design)
There's an experimental design and support for propagating Spans between Actors (without any lifetime handling, which means the user is responsible for finishing the Spans). For this to work, classes must inherit from TracedAbstractActor instead of AbstractActor, and messages must be wrapped using TracedMessage.wrap():
class MyActor extends TracedAbstractActor {
@Override
public Receive createReceive() {
return receiveBuilder()
.match(String.class, msg -> {
// the Span 'foo' will be active for this block,
// and will NOT be finished upon deactivation.
getSender().tell("ciao", getSelf());
})
.build();
}
}
try (Scope scope = tracer.buildSpan("foo").startActive()) {
// scope.span() will be captured as part of TracedMessage.wrap(),
// and MyActor will receive the original 'myMessageObj` instance.
Future<String> f = ask(myActorRef, TracedMessage.wrap("hello"), timeout);
...
}
By default, TracedAbstractActor/TracedMessage use io.opentracing.util.GlobalTracer to activate and fetch the Span respectively, but it's possible to manually specify both the Tracer used to activate and the captured Span:
class MyActor extends TracedAbstractActor {
public Receive createReceive() {
return receiveBuilder()
.match(String.class, msg -> {
// TracedAbstractActor.tracer() returns the Tracer being used,
// either GlobalTracer or the explicit set one.
if (tracer().activeSpan() != null) {
// Use the active Span, to set tags, create children, finish it, etc.
tracer().activeSpan.finish();
}
...
})
.build();
}
}
Span span = tracer.buildSpan("foo").start();
Future<String> f = ask(myActorRef, TracedMessage.wrap(span, "hello"), timeout);
