AWS Lambda

SigNoz supports multiple ways to monitor AWS Lambda. This page helps you choose the right approach for logs, traces, and metrics, then links to detailed setup guides.

Quick Decision Guide

Logs: OTel Extension vs S3

OTel Collector Extension Layer

The OTel Collector runs as a Lambda extension and registers with Lambda's Telemetry API to receive logs automatically. No function code changes required.

Benefits:

• ✅ No code changes to your Lambda function
• ✅ Direct integration via Lambda Telemetry API — no CloudWatch required
• ✅ Lower latency than S3 for high-traffic functions
• ✅ Minimal AWS cost (Lambda extension overhead only)

Trade-offs:

• ❌ Requires adding a Lambda layer and a collector config file
• ❌ Logs are buffered — sent at the start of the next invocation or during function shutdown, not immediately after each invocation. For short-lived or low-traffic functions, expect visible lag.

→ Logs via OTel Extension Layer

S3 Forwarder

Lambda logs flow through CloudWatch Logs to an S3 bucket. A forwarder Lambda function is triggered on new objects and sends them to SigNoz. Use this when you already have a CloudWatch → S3 log archival pipeline.

Benefits:

• ✅ No code changes to your Lambda function
• ✅ Reuses existing S3 log storage and pipelines
• ✅ Decoupled from function execution — delivery failures don't affect your function

Trade-offs:

• ❌ Higher latency — delivery depends on CloudWatch → S3 export interval
• ❌ Additional AWS cost for S3 storage and forwarder Lambda invocations
• ❌ Requires setting up a separate forwarder Lambda function

→ Logs via S3

Traces: OTel Auto-Instrumentation

Add language-specific OTel layers to your Lambda function to automatically capture traces without code changes.

Benefits:

• ✅ No code changes — instrumentation happens via Lambda layers
• ✅ Automatically captures invocation spans and outbound HTTP calls
• ✅ Forwards traces through a separate OTel Collector layer — function response time is not blocked

Trade-offs:

• ❌ Adds cold start latency from layer initialization
• ❌ Custom spans require manual instrumentation
• ❌ Trace-log correlation requires additional setup (OTel SDK logging in your function code)

Supported languages: Python, Node.js, Java, Ruby

→ APM & Tracing Guide

Metrics: CloudWatch Exporter

Scrape Lambda metrics (Invocations, Errors, Duration, Throttles) from CloudWatch using the Prometheus CloudWatch Exporter.

Benefits:

• ✅ No changes to Lambda functions
• ✅ Covers standard Lambda metrics out of the box
• ✅ Minimal AWS cost — reads from CloudWatch, no additional data streams

Trade-offs:

• ❌ Polling-based — not real-time; metrics lag by the scrape interval
• ❌ Requires running a CloudWatch Exporter instance

→ Metrics Guide

One-Click vs Manual

See the full One-Click vs Manual comparison for details.

Combining Approaches

Many teams use a mix:

• One-Click for metrics and pre-built dashboards
• OTel Extension Layer for lower-latency logs
• OTel Auto-Instrumentation for distributed traces

All data flows into the same SigNoz instance and can be queried together.