langfuse-prod-checklist

Langfuse Production Checklist

Contents

• Overview
• Prerequisites
• Instructions
• Output
• Error Handling
• Examples
• Resources

Overview

Comprehensive checklist for deploying Langfuse observability to production with verified configuration, error handling, graceful shutdown, and monitoring.

Prerequisites

• Completed development and staging testing
• Access to production secrets management
• Production Langfuse project created

Instructions

Step 1: Verify Authentication & Security

Confirm production API keys are created, stored in a secrets manager, NOT in code/git, and isolated per environment. Verify PII scrubbing is enabled.

Step 2: Configure SDK Settings

Set flushAt (15-50), flushInterval (5-10s), requestTimeout (10-30s), enabled (true for prod). Use singleton client pattern.

Step 3: Implement Production Error Handling

Wrap all trace operations in try/catch. Call .end() in finally blocks. Log errors to Langfuse with ERROR level. Register shutdownAsync() on SIGTERM/SIGINT.

Step 4: Run Pre-Deployment Verification

Validate env vars are set, API keys are production keys (prefix pk-lf-), trace creation works, and graceful shutdown completes.

Step 5: Configure Production Monitoring

Expose metrics endpoint tracking traces created/failed, flush latency, and error rate.

See detailed implementation for advanced patterns.

Output

• Verified production configuration
• Error handling with graceful degradation
• Graceful shutdown on process signals
• Metrics endpoint for monitoring
• All checklist items validated

Error Handling

Issue	Cause	Solution
Missing traces in prod	Flush not called	Verify shutdown handler
High latency	Large batches	Reduce flushAt
Memory growth	Client recreation	Use singleton pattern
Lost traces on deploy	No graceful shutdown	Add SIGTERM handler

Examples

Production Checklist Summary

Category	Key Items
Auth & Security	Prod keys, secrets manager, env isolation, PII scrubbing
SDK Config	flushAt=25, flushInterval=5000, singleton client
Error Handling	try/catch, .end() in finally, ERROR level traces
Performance	Async tracing, batching, no memory leaks
Observability	Trace URL logging, SDK error monitoring, alerts

Recommended Production Config

{
  flushAt: 25,
  flushInterval: 5000,  # 5000: 5 seconds in ms
  requestTimeout: 15000,  # 15000 = configured value
  enabled: process.env.NODE_ENV === "production",
  debug: false,
}

Resources

• Langfuse Production Guide
• Langfuse Self-Hosting
• Langfuse SDK Reference