Lambda Web Adapter

문서 > 서버리스 > AWS Lambda용 서버리스 모니터링 > Lambda Web Adapter

이 페이지는 아직 영어로 제공되지 않습니다. 번역 작업 중입니다.
현재 번역 프로젝트에 대한 질문이나 피드백이 있으신 경우 언제든지 연락주시기 바랍니다.

Join the Preview!

Lambda Web Adapter is in Preview.

Overview

The AWS Lambda Web Adapter is a framework that allows developers to run web applications on AWS Lambda.

Datadog supports Lambda Web Adapter for Node.js and Python runtimes, therefore providing you a solution to monitoring your web applications running on AWS Lambda.

How to integrate Datadog

The Lambda Web Adapter can run Lambda functions that are packaged as docker images or as Zip files. The following steps are required for instrumenting the Lambda Web Adapter with Datadog for both formats:

Add the Lambda Web Adapter and Datadog extension
Set the required Datadog environment variables
Ignore tracing for the readiness endpoint (*required only when using Datadog tracing)

1. Add the Lambda Web Adapter and Datadog extension

This configuration requires Datadog Lambda extension v77+ and Lambda Web Adapter v0.9.1+. See a sample of the container deployment configuration on Github.

COPY --from=public.ecr.aws/datadog/lambda-extension:77 /opt/. /opt/
COPY --from=public.ecr.aws/awsguru/aws-lambda-adapter:0.9.1 /lambda-adapter /opt/extensions/lambda-adapter

This requires Datadog Lambda extension v77+ and v25+ or later for Lambda Web Adapter. See a sample of the Zip deployment configuration on Github.

for x86

arn:aws:lambda:${AWS::Region}:464622532012:layer:Datadog-Extension:77
arn:aws:lambda:${AWS::Region}:753240598075:layer:LambdaAdapterLayerX86:25

for ARM

arn:aws:lambda:${AWS::Region}:464622532012:layer:Datadog-Extension-ARM:77
arn:aws:lambda:${AWS::Region}:753240598075:layer:LambdaAdapterLayerArm64:25

2. Set the required Datadog environment variables

Since there is no explicit shutdown system for Lambda, traces must be flushed as soon as possible so they are not lost when the Lambda runtime environment is “frozen”.

The transparent tracing requires the Datadog extension to also proxy requests before the Lambda Web Adapter, so the AWS_LWA_LAMBDA_RUNTIME_API_PROXY must be set to allow that. The port can be set with any available port.

DD_TRACE_PARTIAL_FLUSH_MIN_SPANS=1
DD_TRACE_PARTIAL_FLUSH_ENABLED=false

AWS_LWA_LAMBDA_RUNTIME_API_PROXY=127.0.0.1:9002

DD_API_KEY=$YOUR_API_KEY
DD_SERVICE=$YOUR_SERVICE_NAME

Since there is no explicit shutdown system for Lambda, traces must be flushed as soon as possible so they are not lost when the Lambda runtime environment is “frozen”.

DD_TRACE_PARTIAL_FLUSH_MIN_SPANS=1
DD_TRACE_PARTIAL_FLUSH_ENABLED=false

AWS_LWA_LAMBDA_RUNTIME_API_PROXY=127.0.0.1:9002

AWS_LAMBDA_EXEC_WRAPPER=/opt/bootstrap

DD_API_KEY=$YOUR_API_KEY
DD_SERVICE=$YOUR_SERVICE_NAME

3. Ignore tracing for the readiness endpoint

*This step is required only when using Datadog tracing.

Since the Lambda Web Adapter sends readiness check requests once it is loaded, the Datadog extension must not link them to the request that triggered the Lambda function.

The configuration differs depending on the runtime. Assuming the readiness endpoint is the default (GET at /):

NodeJs

const tracer = require('dd-trace').init();
tracer.use('http', {
    blocklist: ['/']
});

Python

import ddtrace

ddtrace.patch_all()
from ddtrace.trace import tracer, TraceFilter
...
class IgnoreEndpointFilter(TraceFilter):
    def __init__(self, pattern, method):
        self.pattern = re.compile(pattern)
        self.method = method

    def process_trace(self, trace):
        for span in trace:
            url = span.get_tag("http.url")
            if (
                url is not None
                and self.pattern.match(url)
                and self.method == span.get_tag("http.method")
            ):
                return None
        return trace


tracer.configure(
    trace_processors=[IgnoreEndpointFilter(r"http://127.0.0.1:8080/", "GET")]
)

Since the Lambda Web Adapter sends readiness check requests once it is loaded, the Datadog extension must not link them to the request that triggered the Lambda function.

The configuration differs depending on the runtime. Assuming the readiness endpoint is the default (GET at /):

NodeJs

const tracer = require('dd-trace').init();
tracer.use('http', {
    blocklist: ['/']
});

Python

import ddtrace

ddtrace.patch_all()
from ddtrace.trace import tracer, TraceFilter
...
class IgnoreEndpointFilter(TraceFilter):
    def __init__(self, pattern, method):
        self.pattern = re.compile(pattern)
        self.method = method

    def process_trace(self, trace):
        for span in trace:
            url = span.get_tag("http.url")
            if (
                url is not None
                and self.pattern.match(url)
                and self.method == span.get_tag("http.method")
            ):
                return None
        return trace


tracer.configure(
    trace_processors=[IgnoreEndpointFilter(r"http://127.0.0.1:8080/", "GET")]
)