Langfuse migration deep dive

Langfuse Migration Deep Dive

Current State

!npm list langfuse @langfuse/client 2>/dev/null | head -5 || echo 'No langfuse packages'

Overview

Comprehensive guide for complex migrations: cloud-to-self-hosted, LangSmith-to-Langfuse, cross-instance data migration, and zero-downtime dual-write patterns.

Prerequisites

• Understanding of source and target Langfuse instances
• API keys for both source and target
• Git branch for migration work
• Rollback plan documented

Migration Scenarios

Scenario	Complexity	Downtime	Data Loss Risk
Cloud to Cloud (different project)	Low	None	None
Cloud to Self-hosted	Medium	Minutes	Low
Self-hosted to Cloud	Medium	Minutes	Low
LangSmith to Langfuse	High	Hours	Medium
SDK v3 to v4+ (no data migration)	Low	None	None

Instructions

Step 1: Export Data from Source Instance

// scripts/export-langfuse.ts
import { LangfuseClient } from "@langfuse/client";
import { writeFileSync, mkdirSync } from "fs";

const source = new LangfuseClient({
  publicKey: process.env.SOURCE_LANGFUSE_PUBLIC_KEY,
  secretKey: process.env.SOURCE_LANGFUSE_SECRET_KEY,
  baseUrl: process.env.SOURCE_LANGFUSE_BASE_URL,
});

async function exportAll(outputDir: string) {
  mkdirSync(outputDir, { recursive: true });

  // Export traces
  let page = 1;
  let allTraces: any[] = [];
  let hasMore = true;

  console.log("Exporting traces...");
  while (hasMore) {
    const result = await source.api.traces.list({ limit: 100, page });
    allTraces.push(...result.data);
    hasMore = result.data.length === 100;
    page++;
    await new Promise((r) => setTimeout(r, 200)); // Rate limit
  }
  writeFileSync(`${outputDir}/traces.json`, JSON.stringify(allTraces, null, 2));
  console.log(`  Exported ${allTraces.length} traces`);

  // Export scores
  page = 1;
  let allScores: any[] = [];
  hasMore = true;

  console.log("Exporting scores...");
  while (hasMore) {
    const result = await source.api.scores.list({ limit: 100, page });
    allScores.push(...result.data);
    hasMore = result.data.length === 100;
    page++;
    await new Promise((r) => setTimeout(r, 200));
  }
  writeFileSync(`${outputDir}/scores.json`, JSON.stringify(allScores, null, 2));
  console.log(`  Exported ${allScores.length} scores`);

  // Export prompts
  console.log("Exporting prompts...");
  const prompts = await source.api.prompts.list({ limit: 100 });
  writeFileSync(`${outputDir}/prompts.json`, JSON.stringify(prompts.data, null, 2));
  console.log(`  Exported ${prompts.data.length} prompts`);

  // Export datasets
  console.log("Exporting datasets...");
  const datasets = await source.api.datasets.list({ limit: 100 });
  const fullDatasets = [];
  for (const ds of datasets.data) {
    const items = await source.api.datasetItems.list({ datasetName: ds.name, limit: 1000 });
    fullDatasets.push({ ...ds, items: items.data });
    await new Promise((r) => setTimeout(r, 200));
  }
  writeFileSync(`${outputDir}/datasets.json`, JSON.stringify(fullDatasets, null, 2));
  console.log(`  Exported ${fullDatasets.length} datasets`);
}

exportAll("./migration-export");

Step 2: Import Data to Target Instance

// scripts/import-langfuse.ts
import { LangfuseClient } from "@langfuse/client";
import { readFileSync } from "fs";

const target = new LangfuseClient({
  publicKey: process.env.TARGET_LANGFUSE_PUBLIC_KEY,
  secretKey: process.env.TARGET_LANGFUSE_SECRET_KEY,
  baseUrl: process.env.TARGET_LANGFUSE_BASE_URL,
});

async function importAll(inputDir: string) {
  // Import prompts first (no dependencies)
  console.log("Importing prompts...");
  const prompts = JSON.parse(readFileSync(`${inputDir}/prompts.json`, "utf-8"));
  for (const prompt of prompts) {
    await target.api.prompts.create({
      name: prompt.name,
      prompt: prompt.prompt,
      type: prompt.type,
      config: prompt.config,
      labels: prompt.labels,
    });
    console.log(`  Imported prompt: ${prompt.name}`);
    await new Promise((r) => setTimeout(r, 100));
  }

  // Import datasets
  console.log("Importing datasets...");
  const datasets = JSON.parse(readFileSync(`${inputDir}/datasets.json`, "utf-8"));
  for (const ds of datasets) {
    await target.api.datasets.create({
      name: ds.name,
      description: ds.description,
      metadata: { ...ds.metadata, migratedFrom: "source-instance" },
    });

    for (const item of ds.items || []) {
      await target.api.datasetItems.create({
        datasetName: ds.name,
        input: item.input,
        expectedOutput: item.expectedOutput,
        metadata: item.metadata,
      });
      await new Promise((r) => setTimeout(r, 50));
    }
    console.log(`  Imported dataset: ${ds.name} (${ds.items?.length || 0} items)`);
  }

  console.log("Import complete.");
  console.log("Note: Traces and scores are historical -- they reference old trace IDs.");
  console.log("New traces will be created by your application pointing to the target.");
}

importAll("./migration-export");

Step 3: Dual-Write for Zero-Downtime Migration

Write traces to both instances during transition:

// src/lib/dual-write-langfuse.ts
import { LangfuseSpanProcessor } from "@langfuse/otel";
import { NodeSDK } from "@opentelemetry/sdk-node";

// Create processors for both instances
const sourceProcessor = new LangfuseSpanProcessor({
  publicKey: process.env.SOURCE_LANGFUSE_PUBLIC_KEY,
  secretKey: process.env.SOURCE_LANGFUSE_SECRET_KEY,
  baseUrl: process.env.SOURCE_LANGFUSE_BASE_URL,
});

const targetProcessor = new LangfuseSpanProcessor({
  publicKey: process.env.TARGET_LANGFUSE_PUBLIC_KEY,
  secretKey: process.env.TARGET_LANGFUSE_SECRET_KEY,
  baseUrl: process.env.TARGET_LANGFUSE_BASE_URL,
});

// Both processors receive all spans
const sdk = new NodeSDK({
  spanProcessors: [sourceProcessor, targetProcessor],
});
sdk.start();

// Migration timeline:
// Week 1: Dual-write enabled, verify target receives traces
// Week 2: Validate data parity between instances
// Week 3: Switch primary to target, keep source as backup
// Week 4: Remove source processor

Step 4: Validate Migration

// scripts/validate-migration.ts
import { LangfuseClient } from "@langfuse/client";

const source = new LangfuseClient({
  publicKey: process.env.SOURCE_LANGFUSE_PUBLIC_KEY,
  secretKey: process.env.SOURCE_LANGFUSE_SECRET_KEY,
  baseUrl: process.env.SOURCE_LANGFUSE_BASE_URL,
});

const target = new LangfuseClient({
  publicKey: process.env.TARGET_LANGFUSE_PUBLIC_KEY,
  secretKey: process.env.TARGET_LANGFUSE_SECRET_KEY,
  baseUrl: process.env.TARGET_LANGFUSE_BASE_URL,
});

async function validate() {
  // Compare prompt counts
  const sourcePrompts = await source.api.prompts.list({ limit: 100 });
  const targetPrompts = await target.api.prompts.list({ limit: 100 });
  console.log(`Prompts: source=${sourcePrompts.data.length}, target=${targetPrompts.data.length}`);

  // Compare dataset counts
  const sourceDatasets = await source.api.datasets.list({ limit: 100 });
  const targetDatasets = await target.api.datasets.list({ limit: 100 });
  console.log(`Datasets: source=${sourceDatasets.data.length}, target=${targetDatasets.data.length}`);

  // Compare recent trace counts (dual-write period)
  const since = new Date(Date.now() - 3600000).toISOString();
  const sourceTraces = await source.api.traces.list({ fromTimestamp: since, limit: 100 });
  const targetTraces = await target.api.traces.list({ fromTimestamp: since, limit: 100 });
  console.log(`Recent traces (1h): source=${sourceTraces.data.length}, target=${targetTraces.data.length}`);

  const variance = Math.abs(sourceTraces.data.length - targetTraces.data.length) / Math.max(sourceTraces.data.length, 1);
  console.log(`Trace variance: ${(variance * 100).toFixed(1)}% (target: <5%)`);
}

validate();

Step 5: Cutover and Cleanup

// After validation passes:

// 1. Update environment variables to point to target
// LANGFUSE_PUBLIC_KEY=pk-lf-target-...
// LANGFUSE_SECRET_KEY=sk-lf-target-...
// LANGFUSE_BASE_URL=https://target.langfuse.com

// 2. Remove dual-write (use single processor)
const sdk = new NodeSDK({
  spanProcessors: [targetProcessor], // Only target
});

// 3. Keep source instance running for 30 days (rollback window)
// 4. After 30 days, decommission source

Rollback Plan

set -euo pipefail
# If migration fails, switch back to source:

# 1. Update environment variables
export LANGFUSE_PUBLIC_KEY="pk-lf-source-..."
export LANGFUSE_SECRET_KEY="sk-lf-source-..."
export LANGFUSE_BASE_URL="https://source.langfuse.com"

# 2. Restart application
# 3. Verify traces flowing to source

Error Handling

Issue	Cause	Solution
Export timeout	Too much data	Paginate with smaller page sizes
Import duplicates	Re-running import	Use idempotent creates with unique names
Dual-write divergence	One instance failing	Monitor both, alert on variance > 5%
Missing prompts	Not exported	Export prompts before datasets

Resources

• Self-Hosting Guide
• Upgrade Guide
• API Reference
• v3 to v4 Migration