Sending trace data to Amazon X-Ray

You can send trace data to X-Ray in the form of segment documents. A segment document is a JSON formatted string that contains information about the work that your application does in service of a request. Your application can record data about the work that it does itself in segments, or work that uses downstream services and resources in subsegments.

Segments record information about the work that your application does. A segment, at a minimum, records the time spent on a task, a name, and two IDs. The trace ID tracks the request as it travels between services. The segment ID tracks the work done for the request by a single service.

{
  "name" : "Scorekeep",
  "id" : "70de5b6f19ff9a0a",
  "start_time" : 1.478293361271E9,
  "trace_id" : "1-581cf771-a006649127e371903a2de979",
  "end_time" : 1.478293361449E9
}

When a request is received, you can send an in-progress segment as a placeholder until the request is completed.

{
  "name" : "Scorekeep",
  "id" : "70de5b6f19ff9a0b",
  "start_time" : 1.478293361271E9,
  "trace_id" : "1-581cf771-a006649127e371903a2de979",
  “in_progress”: true
}

You can send segments to X-Ray directly, with PutTraceSegments, or through the X-Ray daemon.

Most applications call other services or access resources with the Amazon SDK. Record information about downstream calls in subsegments. X-Ray uses subsegments to identify downstream services that don't send segments and create entries for them on the service graph.

A subsegment can be embedded in a full segment document, or sent separately. Send subsegments separately to asynchronously trace downstream calls for long-running requests, or to avoid exceeding the maximum segment document size (64 kB).

{
  "name" : "www2.example.com",
  "id" : "70de5b6f19ff9a0c",
  "start_time" : 1.478293361271E9,
  "trace_id" : "1-581cf771-a006649127e371903a2de979"
  “end_time” : 1.478293361449E9,
  “type” : “subsegment”,
  “parent_id” : “70de5b6f19ff9a0b”
}

For more information on the fields and values that you can include in segments and subsegments, see Amazon X-Ray segment documents.

Generating trace IDs

To send data to X-Ray, you must generate a unique trace ID for each request.

You can write a script to generate X-Ray trace IDs for testing. Here are two examples.

Python

import time
import os
import binascii

START_TIME = time.time()
HEX=hex(int(START_TIME))[2:]
TRACE_ID="1-{}-{}".format(HEX, binascii.hexlify(os.urandom(12)).decode('utf-8'))

Bash

START_TIME=$(date +%s)
HEX_TIME=$(printf '%x\n' $START_TIME)
GUID=$(dd if=/dev/random bs=12 count=1 2>/dev/null | od -An -tx1 | tr -d ' \t\n')
TRACE_ID="1-$HEX_TIME-$GUID"

See the Scorekeep sample application for scripts that create trace IDs and send segments to the X-Ray daemon.

Using PutTraceSegments

You can upload segment documents with the PutTraceSegments API. The API has a single parameter, TraceSegmentDocuments, that takes a list of JSON segment documents.

With the AWS CLI, use the aws xray put-trace-segments command to send segment documents directly to X-Ray.

$ DOC='{"trace_id": "1-5960082b-ab52431b496add878434aa25", "id": "6226467e3f845502", "start_time": 1498082657.37518, "end_time": 1498082695.4042, "name": "test.elasticbeanstalk.com"}'
$ aws xray put-trace-segments --trace-segment-documents "$DOC"
{
    "UnprocessedTraceSegments": []
}

The output lists any segments that failed processing. For example, if the date in the trace ID is too far in the past, you see an error like the following.

{
    "UnprocessedTraceSegments": [
        {
            "ErrorCode": "InvalidTraceId",
            "Message": "Invalid segment. ErrorCode: InvalidTraceId",
            "Id": "6226467e3f845502"
        }
    ]
}

You can pass multiple segment documents at the same time, separated by spaces.

$ aws xray put-trace-segments --trace-segment-documents "$DOC1" "$DOC2"

Sending segment documents to the X-Ray daemon

Instead of sending segment documents to the X-Ray API, you can send segments and subsegments to the X-Ray daemon, which will buffer them and upload to the X-Ray API in batches. The X-Ray SDK sends segment documents to the daemon to avoid making calls to Amazon directly.

Send the segment in JSON over UDP port 2000, prepended by the daemon header, {"format": "json", "version": 1}\n

{"format": "json", "version": 1}\n{"trace_id": "1-5759e988-bd862e3fe1be46a994272793", "id": "defdfd9912dc5a56", "start_time": 1461096053.37518, "end_time": 1461096053.4042, "name": "test.elasticbeanstalk.com"}

On Linux, you can send segment documents to the daemon from a Bash terminal. Save the header and segment document to a text file and pipe it to /dev/udp with cat.

$ cat segment.txt > /dev/udp/127.0.0.1/2000

{"format": "json", "version": 1}
{"trace_id": "1-594aed87-ad72e26896b3f9d3a27054bb", "id": "6226467e3f845502", "start_time": 1498082657.37518, "end_time": 1498082695.4042, "name": "test.elasticbeanstalk.com"}

Check the daemon log to verify that it sent the segment to X-Ray.

2017-07-07T01:57:24Z [Debug] processor: sending partial batch
2017-07-07T01:57:24Z [Debug] processor: segment batch size: 1. capacity: 50
2017-07-07T01:57:24Z [Info] Successfully sent batch of 1 segments (0.020 seconds)