Skip to content

Tracer

Tracer is an opinionated thin wrapper for AWS X-Ray Python SDK.

Tracer showcase

Key features

  • Auto capture cold start as annotation, and responses or full exceptions as metadata
  • Auto-disable when not running in AWS Lambda environment
  • Support tracing async methods, generators, and context managers
  • Auto patch supported modules by AWS X-Ray

Getting started

Permissions

Before your use this utility, your AWS Lambda function must have permissions to send traces to AWS X-Ray.

AWS Serverless Application Model (SAM) example
1
2
3
4
5
6
7
8
9
Resources:
  HelloWorldFunction:
    Type: AWS::Serverless::Function
    Properties:
      Runtime: python3.8
      Tracing: Active
      Environment:
        Variables:
          POWERTOOLS_SERVICE_NAME: example

Lambda handler

You can quickly start by initializing Tracer and use capture_lambda_handler decorator for your Lambda handler.

Tracing Lambda handler with capture_lambda_handler
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
from aws_lambda_powertools import Tracer

tracer = Tracer() # Sets service via env var
# OR tracer = Tracer(service="example")

@tracer.capture_lambda_handler
def handler(event, context):
    charge_id = event.get('charge_id')
    payment = collect_payment(charge_id)
    ...

capture_lambda_handler performs these additional tasks to ease operations:

  • Creates a ColdStart annotation to easily filter traces that have had an initialization overhead
  • Creates a Service annotation if service parameter or POWERTOOLS_SERVICE_NAME is set
  • Captures any response, or full exceptions generated by the handler, and include as tracing metadata

Annotations & Metadata

Annotations are key-values associated with traces and indexed by AWS X-Ray. You can use them to filter traces and to create Trace Groups to slice and dice your transactions.

Adding annotations with put_annotation method
1
2
3
4
5
6
7
from aws_lambda_powertools import Tracer
tracer = Tracer()

@tracer.capture_lambda_handler
def handler(event, context):
    ...
    tracer.put_annotation(key="PaymentStatus", value="SUCCESS")

Metadata are key-values also associated with traces but not indexed by AWS X-Ray. You can use them to add additional context for an operation using any native object.

Adding arbitrary metadata with put_metadata method
1
2
3
4
5
6
7
8
from aws_lambda_powertools import Tracer
tracer = Tracer()

@tracer.capture_lambda_handler
def handler(event, context):
    ...
    ret = some_logic()
    tracer.put_metadata(key="payment_response", value=ret)

Synchronous functions

You can trace synchronous functions using the capture_method decorator.

Tracing an arbitrary function with capture_method
1
2
3
4
5
@tracer.capture_method
def collect_payment(charge_id):
    ret = requests.post(PAYMENT_ENDPOINT) # logic
    tracer.put_annotation("PAYMENT_STATUS", "SUCCESS") # custom annotation
    return ret
Note: Function responses are auto-captured and stored as JSON, by default.

Use capture_response parameter to override this behaviour.

The serialization is performed by aws-xray-sdk via jsonpickle module. This can cause side effects for file-like objects like boto S3 StreamingBody, where its response will be read only once during serialization.

Asynchronous and generator functions

Warning

We do not support asynchronous Lambda handler

You can trace asynchronous functions and generator functions (including context managers) using capture_method.

1
2
3
4
5
6
7
8
9
import asyncio
import contextlib
from aws_lambda_powertools import Tracer

tracer = Tracer()

@tracer.capture_method
async def collect_payment():
    ...

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
import asyncio
import contextlib
from aws_lambda_powertools import Tracer

tracer = Tracer()

@contextlib.contextmanager
@tracer.capture_method
def collect_payment_ctxman():
    yield result
    ...

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
import asyncio
import contextlib
from aws_lambda_powertools import Tracer

tracer = Tracer()

@tracer.capture_method
def collect_payment_gen():
    yield result
    ...

Advanced

Patching modules

Tracer automatically patches all supported libraries by X-Ray during initialization, by default. Underneath, AWS X-Ray SDK checks whether a supported library has been imported before patching.

If you're looking to shave a few microseconds, or milliseconds depending on your function memory configuration, you can patch specific modules using patch_modules param:

Example of explicitly patching boto3 and requests only
1
2
3
4
5
6
7
import boto3
import requests

from aws_lambda_powertools import Tracer

modules_to_be_patched = ["boto3", "requests"]
tracer = Tracer(patch_modules=modules_to_be_patched)

Disabling response auto-capture

Use capture_response=False parameter in both capture_lambda_handler and capture_method decorators to instruct Tracer not to serialize function responses as metadata.

Info: This is useful in three common scenarios
  1. You might return sensitive information you don't want it to be added to your traces
  2. You might manipulate streaming objects that can be read only once; this prevents subsequent calls from being empty
  3. You might return more than 64K of data e.g., message too long error
1
2
3
4
5
6
7
8
9
from aws_lambda_powertools import Tracer

@tracer.capture_method(capture_response=False)
def fetch_sensitive_information():
    return "sensitive_information"

@tracer.capture_lambda_handler(capture_response=False)
def handler(event, context):
    sensitive_information = fetch_sensitive_information()

1
2
3
4
5
6
7
from aws_lambda_powertools import Tracer

@tracer.capture_method(capture_response=False)
def get_s3_object(bucket_name, object_key):
    s3 = boto3.client("s3")
    s3_object = get_object(Bucket=bucket_name, Key=object_key)
    return s3_object

Disabling exception auto-capture

Use capture_error=False parameter in both capture_lambda_handler and capture_method decorators to instruct Tracer not to serialize exceptions as metadata.

Info

Useful when returning sensitive information in exceptions/stack traces you don't control

Disabling exception auto-capture for tracing metadata
1
2
3
4
5
from aws_lambda_powertools import Tracer

@tracer.capture_lambda_handler(capture_error=False)
def handler(event, context):
    raise ValueError("some sensitive info in the stack trace...")

Ignoring certain HTTP endpoints

You might have endpoints you don't want requests to be traced, perhaps due to the volume of calls or sensitive URLs.

You can use ignore_endpoint method with the hostname and/or URLs you'd like it to be ignored - globs (*) are allowed.

Ignoring certain HTTP endpoints from being traced
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
from aws_lambda_powertools import Tracer

tracer = Tracer()
# ignore all calls to `ec2.amazon.com`
tracer.ignore_endpoint(hostname="ec2.amazon.com")
# ignore calls to `*.sensitive.com/password` and  `*.sensitive.com/credit-card`
tracer.ignore_endpoint(hostname="*.sensitive.com", urls=["/password", "/credit-card"])


def ec2_api_calls():
    return "suppress_api_responses"

@tracer.capture_lambda_handler
def handler(event, context):
    for x in long_list:
        ec2_api_calls()

Tracing aiohttp requests

Info

This snippet assumes you have aiohttp as a dependency

You can use aiohttp_trace_config function to create a valid aiohttp trace_config object. This is necessary since X-Ray utilizes aiohttp trace hooks to capture requests end-to-end.

Tracing aiohttp requests
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
import asyncio
import aiohttp

from aws_lambda_powertools import Tracer
from aws_lambda_powertools.tracing import aiohttp_trace_config

tracer = Tracer()

async def aiohttp_task():
    async with aiohttp.ClientSession(trace_configs=[aiohttp_trace_config()]) as session:
        async with session.get("https://httpbin.org/json") as resp:
            resp = await resp.json()
            return resp

Escape hatch mechanism

You can use tracer.provider attribute to access all methods provided by AWS X-Ray xray_recorder object.

This is useful when you need a feature available in X-Ray that is not available in the Tracer utility, for example thread-safe, or context managers.

Tracing a code block with in_subsegment escape hatch
1
2
3
4
5
6
7
8
9
from aws_lambda_powertools import Tracer

tracer = Tracer()

@tracer.capture_lambda_handler
def handler(event, context):
    with tracer.provider.in_subsegment('## custom subsegment') as subsegment:
        ret = some_work()
        subsegment.put_metadata('response', ret)

Concurrent asynchronous functions

Warning

X-Ray SDK will raise an exception when async functions are run and traced concurrently

A safe workaround mechanism is to use in_subsegment_async available via Tracer escape hatch (tracer.provider).

Workaround to safely trace async concurrent functions
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
import asyncio

from aws_lambda_powertools import Tracer
tracer = Tracer()

async def another_async_task():
    async with tracer.provider.in_subsegment_async("## another_async_task") as subsegment:
        subsegment.put_annotation(key="key", value="value")
        subsegment.put_metadata(key="key", value="value", namespace="namespace")
        ...

async def another_async_task_2():
    ...

@tracer.capture_method
async def collect_payment(charge_id):
    asyncio.gather(another_async_task(), another_async_task_2())
    ...

Reusing Tracer across your code

Tracer keeps a copy of its configuration after the first initialization. This is useful for scenarios where you want to use Tracer in more than one location across your code base.

Warning: Import order matters when using Lambda Layers or multiple modules

Do not set auto_patch=False when reusing Tracer in Lambda Layers, or in multiple modules.

This can result in the first Tracer config being inherited by new instances, and their modules not being patched.

Tracer will automatically ignore imported modules that have been patched.

1
2
3
4
5
6
7
8
9
from aws_lambda_powertools import Tracer
from payment import collect_payment

tracer = Tracer(service="payment")

@tracer.capture_lambda_handler
def handler(event, context):
    charge_id = event.get('charge_id')
    payment = collect_payment(charge_id)

A new instance of Tracer will be created but will reuse the previous Tracer instance configuration, similar to a Singleton.

1
2
3
4
5
6
7
from aws_lambda_powertools import Tracer

tracer = Tracer(service="payment")

@tracer.capture_method
def collect_payment(charge_id: str):
    ...

Testing your code

Tracer is disabled by default when not running in the AWS Lambda environment - This means no code changes or environment variables to be set.

Tips

  • Use annotations on key operations to slice and dice traces, create unique views, and create metrics from it via Trace Groups
  • Use a namespace when adding metadata to group data more easily
  • Annotations and metadata are added to the current subsegment opened. If you want them in a specific subsegment, use a context manager via the escape hatch mechanism
Last update: 2021-12-31
Back to top