Overview

Install, configure, scale, and operate self-hosted n8n instances. Targets n8n 2.x (stable 2.43 at the time of writing; 2.0 replaced "activate" with publish/unpublish and made task runners the default). Covers Docker Compose production setups, CLI administration, REST API automation, queue-mode scaling with Redis workers, backup/restore, monitoring, and troubleshooting. For building n8n workflows and configuring nodes, see the n8n-workflow skill.

Instructions

Task A: Production Docker Compose Setup

Basic setup with PostgreSQL (recommended over default SQLite):

yaml

# docker-compose.yml
services:
  postgres:
    image: postgres:16
    restart: always
    environment:
      POSTGRES_USER: ${POSTGRES_USER}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
      POSTGRES_DB: ${POSTGRES_DB}
    volumes:
      - db_storage:/var/lib/postgresql/data
    healthcheck:
      test: ['CMD-SHELL', 'pg_isready -h localhost -U ${POSTGRES_USER} -d ${POSTGRES_DB}']
      interval: 5s
      timeout: 5s
      retries: 10

  n8n:
    image: docker.n8n.io/n8nio/n8n:${N8N_VERSION}
    restart: always
    ports:
      - "5678:5678"
    environment:
      - DB_TYPE=postgresdb
      - DB_POSTGRESDB_HOST=postgres
      - DB_POSTGRESDB_PORT=5432
      - DB_POSTGRESDB_DATABASE=${POSTGRES_DB}
      - DB_POSTGRESDB_USER=${POSTGRES_USER}
      - DB_POSTGRESDB_PASSWORD=${POSTGRES_PASSWORD}
      - N8N_ENCRYPTION_KEY=${ENCRYPTION_KEY}
      - WEBHOOK_URL=${WEBHOOK_URL}
      - GENERIC_TIMEZONE=${GENERIC_TIMEZONE:-UTC}
      - N8N_ENFORCE_SETTINGS_FILE_PERMISSIONS=true
      - N8N_PROXY_HOPS=1            # one reverse proxy in front
      - N8N_RUNNERS_MODE=external   # Code nodes run in the sidecar below
      - N8N_RUNNERS_BROKER_LISTEN_ADDRESS=0.0.0.0
      - N8N_RUNNERS_AUTH_TOKEN=${RUNNERS_AUTH_TOKEN}
    volumes:
      - n8n_storage:/home/node/.n8n
    depends_on:
      postgres:
        condition: service_healthy

  task-runners:
    image: n8nio/runners:${N8N_VERSION}   # must match the n8n image version
    restart: always
    environment:
      - N8N_RUNNERS_TASK_BROKER_URI=http://n8n:5679
      - N8N_RUNNERS_AUTH_TOKEN=${RUNNERS_AUTH_TOKEN}
    depends_on:
      - n8n

volumes:
  db_storage:
  n8n_storage:

Create a .env alongside with POSTGRES_USER, POSTGRES_PASSWORD, POSTGRES_DB, ENCRYPTION_KEY (generate with openssl rand -hex 24), RUNNERS_AUTH_TOKEN (same way), N8N_VERSION (pin the current stable, e.g. 2.43.4, from the n8n releases page), WEBHOOK_URL, and GENERIC_TIMEZONE. Start with docker compose up -d.

Task runners isolate Code-node JavaScript and Python from n8n itself. Internal mode (the default child process) shares n8n's user and is deprecated; in production use external mode with the n8nio/runners sidecar as above. N8N_RUNNERS_ENABLED is deprecated since 2.0 and not needed. From n8n 2.30 N8N_WEBHOOK_URL is the preferred name of WEBHOOK_URL.

Task B: Scale with Queue Mode (Redis + Workers)

For high-volume workloads, add Redis and worker processes:

yaml

# Add to docker-compose.yml
x-n8n-shared: &n8n-shared
  image: docker.n8n.io/n8nio/n8n:${N8N_VERSION}
  restart: always
  environment:
    - DB_TYPE=postgresdb
    - DB_POSTGRESDB_HOST=postgres
    - DB_POSTGRESDB_DATABASE=${POSTGRES_DB}
    - DB_POSTGRESDB_USER=${POSTGRES_USER}
    - DB_POSTGRESDB_PASSWORD=${POSTGRES_PASSWORD}
    - EXECUTIONS_MODE=queue
    - QUEUE_BULL_REDIS_HOST=redis
    - QUEUE_HEALTH_CHECK_ACTIVE=true
    - N8N_ENCRYPTION_KEY=${ENCRYPTION_KEY}
  volumes:
    - n8n_storage:/home/node/.n8n
  depends_on:
    postgres:
      condition: service_healthy
    redis:
      condition: service_healthy

services:
  redis:
    image: redis:7-alpine
    restart: always
    volumes:
      - redis_storage:/data
    healthcheck:
      test: ['CMD', 'redis-cli', 'ping']
      interval: 5s

  n8n:
    <<: *n8n-shared
    ports:
      - "5678:5678"

  n8n-worker:
    <<: *n8n-shared
    command: worker
    # Scale with: docker compose up -d --scale n8n-worker=3

Each worker needs its own task-runner sidecar in external mode. Health endpoints on workers (/healthz, /healthz/readiness) exist only with QUEUE_HEALTH_CHECK_ACTIVE=true. Queue mode needs PostgreSQL; SQLite is not supported with workers. Optional webhook processors (command: webhook) scale incoming webhooks: route /webhook/* and /webhook-waiting/* to them and keep /webhook-test/* and the editor on the main instance.

Task C: CLI Administration

Run CLI commands inside Docker: docker exec -u node -it n8n n8n <command>.

bash

# Export/import workflows (use a separate directory per entity type)
n8n export:workflow --backup --output=backups/workflows/
n8n export:workflow --id=Hk3p9WqLm2xYtRvB --published --output=workflow.json   # published version
n8n import:workflow --input=workflow.json
n8n import:workflow --separate --input=backups/workflows/   # imported workflows come in unpublished

# Export/import credentials (--decrypted for cross-instance migration)
n8n export:credentials --backup --output=backups/credentials/
n8n export:credentials --all --decrypted --output=creds.json
n8n import:credentials --separate --input=backups/credentials/

# Full database backup/restore
n8n export:entities --outputDir=./backup --includeExecutionHistoryDataTables=true
n8n import:entities --inputDir=./backup --truncateTables=true

n8n execute --id=Hk3p9WqLm2xYtRvB   # run a saved workflow

# Publish/unpublish workflows (2.0+; update:workflow --active is deprecated).
# These edit the database: restart n8n for them to take effect.
n8n publish:workflow --id=Hk3p9WqLm2xYtRvB
n8n unpublish:workflow --id=Hk3p9WqLm2xYtRvB
n8n unpublish:workflow --all     # publish has no --all, by design

# User management
n8n user-management:reset          # Reset to setup state (forgot password)
n8n mfa:disable --email=user@x.com # Disable MFA for locked-out user

n8n audit               # security audit

--backup exports only workflows and credentials, not users, variables, execution history or the encryption key. A complete backup also needs the .n8n folder (it holds the key) and a PostgreSQL dump. --activeState=fromJson on import restores published state, but only in queue or multi-main mode.

Task D: REST API Automation

Create an API key in the n8n settings and send it in the X-N8N-API-KEY header. The self-hosted playground is at /api/v1/docs; the base path is /api/v1.

bash

# List all workflows
curl -s -H "X-N8N-API-KEY: $N8N_API_KEY" http://localhost:5678/api/v1/workflows | jq '.data[].name'

# Publish a workflow (replaces /activate, deprecated)
curl -X POST -H "X-N8N-API-KEY: $N8N_API_KEY" http://localhost:5678/api/v1/workflows/Hk3p9WqLm2xYtRvB/publish

# Unpublish a workflow (replaces /deactivate, deprecated)
curl -X POST -H "X-N8N-API-KEY: $N8N_API_KEY" http://localhost:5678/api/v1/workflows/Hk3p9WqLm2xYtRvB/unpublish

# List failed executions
curl -s -H "X-N8N-API-KEY: $N8N_API_KEY" "http://localhost:5678/api/v1/executions?status=error" | jq '.data[] | {id, workflowId, status}'

Lists are paginated: pass the nextCursor value back as cursor (limit max 250).

Task E: Key Environment Variables

Variable	Default	Purpose
`DB_TYPE`	`sqlite`	`sqlite` or `postgresdb`
`N8N_ENCRYPTION_KEY`	random	Credential encryption (preserve across restarts)
`WEBHOOK_URL`	auto	Public URL for webhooks behind reverse proxy
`EXECUTIONS_MODE`	`regular`	`regular` or `queue` (for scaling)
`EXECUTIONS_DATA_PRUNE`	`true`	Auto-prune old execution data
`N8N_METRICS`	`false`	Enable Prometheus `/metrics` endpoint
`N8N_RUNNERS_MODE`	`internal`	`external` runs Code nodes in the `n8nio/runners` sidecar (recommended in production)
`GENERIC_TIMEZONE`	`America/New_York`	Timezone for schedule triggers

Use _FILE suffix for Docker Secrets (e.g., DB_POSTGRESDB_PASSWORD_FILE=/run/secrets/db_pass).

Examples

Example 1: Deploy n8n with PostgreSQL in production

User request: "Set up n8n in production with Docker Compose"

Actions taken:

bash

$ mkdir n8n-production && cd n8n-production
$ openssl rand -hex 24   # use for ENCRYPTION_KEY and RUNNERS_AUTH_TOKEN
> 9f2c61b7d0e84a35c1f7a3e92b64d85077ac13e6f2b9d4c8

# Create .env with POSTGRES_USER, POSTGRES_PASSWORD, POSTGRES_DB, ENCRYPTION_KEY,
# RUNNERS_AUTH_TOKEN, N8N_VERSION=2.43.4, WEBHOOK_URL, GENERIC_TIMEZONE

# Create docker-compose.yml (Task A template)

$ docker compose up -d
> [+] Running 4/4
>  ✔ Network n8n-production_default       Created
>  ✔ Container n8n-production-postgres-1  Healthy
>  ✔ Container n8n-production-n8n-1       Started
>  ✔ Container n8n-production-task-runners-1  Started

$ docker compose logs n8n --tail 5
> n8n  | Editor is now accessible via: http://localhost:5678

Example 2: Back up and restore workflows

User request: "Back up all my n8n workflows and credentials"

Actions taken:

bash

$ docker exec -u node n8n n8n export:workflow --backup --output=/home/node/.n8n/backups/workflows/
> Successfully exported 12 workflows
$ docker exec -u node n8n n8n export:credentials --backup --output=/home/node/.n8n/backups/credentials/
> Successfully exported 8 credentials

$ docker cp n8n:/home/node/.n8n/backups/ ./n8n-backups/

# Restore on a new instance (finish the owner setup in the UI first)
$ docker cp ./n8n-backups/ n8n-new:/home/node/.n8n/backups/
$ docker exec -u node n8n-new n8n import:workflow --separate --input=/home/node/.n8n/backups/workflows/
> Successfully imported 12 workflows
$ docker exec -u node n8n-new n8n import:credentials --separate --input=/home/node/.n8n/backups/credentials/

The new instance needs the same N8N_ENCRYPTION_KEY, or export credentials with --decrypted and delete that file after the import. Imported workflows arrive unpublished; publish them in the UI.

Example 3: Recover from a failed update

User request: "My n8n instance is crashing after update"

Actions taken:

bash

$ docker compose logs n8n --tail 30
> Error: Migration "AddWorkflowVersioning" failed ...

# Roll back: stop, pin the previous image tag in .env (N8N_VERSION=2.42.0),
# restore the PostgreSQL dump and the .n8n volume taken before the update
$ docker compose down
$ pg_restore --clean -d n8n backups/n8n-before-update.dump   # run against the postgres container
$ docker compose up -d
> n8n  | Editor is now accessible via: http://localhost:5678

Pinning an older image does not undo database migrations, so take a database dump and a copy of the .n8n volume before every upgrade.

Guidelines

Always set N8N_ENCRYPTION_KEY explicitly and back it up; without it credentials cannot be decrypted on a new instance.
Use PostgreSQL for anything beyond a few workflows; queue mode requires it.
Set WEBHOOK_URL when behind a reverse proxy — webhooks fail silently without it.
Use n8n export:credentials --decrypted when migrating between instances with different encryption keys, and delete the plain-text file afterwards.
Pin the image to a version (2.43.4), not latest; the stable tag is also fine for production but changes weekly. Keep the n8nio/runners tag equal to the n8n tag.
Enable execution pruning (EXECUTIONS_DATA_PRUNE=true) to prevent database bloat.
In queue mode, all instances must share the same N8N_ENCRYPTION_KEY and PostgreSQL database.
For health checks, use /healthz (instance reachable) and /healthz/readiness (DB connected and migrated). /metrics needs N8N_METRICS=true.
Code nodes without external task runners can reach credentials and the encryption key; do not let untrusted users edit workflows on such an instance.

n8n Self-Host

Validation

Prove n8n Self-Host on your task

Getting Started

Example Prompts

Documentation

Overview

Instructions

Task A: Production Docker Compose Setup

Task B: Scale with Queue Mode (Redis + Workers)

Task C: CLI Administration

Task D: REST API Automation

Task E: Key Environment Variables

Examples

Example 1: Deploy n8n with PostgreSQL in production

Example 2: Back up and restore workflows

Example 3: Recover from a failed update

Guidelines

Information

Use Cases