x-ray-instrument-custom-namespace.js

// This handler function gets called for each SQS message that arrives in my process
mySqsHandler = function(msg) {
  var self = this;

  if (!msg.Attributes || !msg.Attributes.AWSTraceHeader) {
    // This message was not selected for tracing by the sampler in
    // a higher level service
    runMySqsLogic(msg, function() {
      // Noop, no segment to run
    });
  }

  var trace = AWSXRay.utils.processTraceData(msg.Attributes.AWSTraceHeader);
  var segment = new AWSXRay.Segment('message-distributor', trace.Root, trace.Parent);
  segment.notTraced = (trace.Sampled !== '1');

  // Create an instrumented namespace that will create trace spans.
  var ns = AWSXRay.getNamespace();
  ns.run(function() {
    AWSXRay.setSegment(segment);
    runMySqsLogic(msg, function() {
      // After my SQS logic ran don't forget to close the segment.
      segment.close();
    });
  });
};

@nathanpeck I'm following from your tweeter feedback, and what I'm missing here is the AWSTraceHeader, which I don't really have in the context of our code. What should be the proper way to start a new trace from a "blank slate"?

I'm specifically referring to lines 13-14 where your extract de traceHeader from the msg, data that we don't have :(

Ah in that case you can skip that. For this example the only way the SQS message would get to this process was because it was generated by a higher level service that would have initiated the transaction and have a trace header already. You can start a brand new segment from scratch if you want to by just leaving out that if statement on lines 5-11

ok, and what about testing this locally on my computer, how would the x-ray agent should be run?