Scaling Self-Hosted n8n to 100k Daily Executions with Kubernetes and Redis Queues

When you first install n8n using docker run -p 5678:5678 n8nio/n8n, it runs in single-process mode with an embedded SQLite database.

It is brilliant for local testing.

If you point a production webhook receiving 10,000 requests per hour at that single container, the Node.js event loop will freeze, memory will hit 100%, and incoming customer webhooks will return HTTP 504 Gateway Timeouts.

Here is the Enterprise High-Availability (HA) n8n Architecture I deploy on Kubernetes to comfortably process over 100,000 workflow executions per day.

The Scaled Queue Mode Topology

                       [Ingress NGINX Load Balancer]
                                    │
                  ┌─────────────────┴─────────────────┐
                  ▼                                   ▼
    [n8n Webhook Replicas (x3)]             [n8n Editor UI (x1)]
                  │                                   │
                  └─────────────────┬─────────────────┘
                                    ▼
                         [Redis Cluster (Bull Queue)]
                                    │
                  ┌─────────────────┼─────────────────┐
                  ▼                 ▼                 ▼
          [Worker Pod 1]     [Worker Pod 2]    [Worker Pod 3]
                  │                 │                 │
                  └─────────────────┬─────────────────┘
                                    ▼
                   [PostgreSQL Primary + Read Replica]

Step 1: Environment Parameters for Queue Mode

In your Kubernetes Deployment manifest, configure n8n to run in Queue Mode with Redis as the broker:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: n8n-worker
spec:
  replicas: 4
  template:
    spec:
      containers:
      - name: n8n-worker
        image: n8nio/n8n:latest
        args: ["worker"]
        env:
        - name: EXECUTIONS_MODE
          value: "queue"
        - name: QUEUE_BULL_REDIS_HOST
          value: "redis-cluster.internal"
        - name: DB_TYPE
          value: "postgresdb"
        - name: DB_POSTGRESDB_HOST
          value: "postgres-primary.internal"
        - name: EXECUTIONS_DATA_PRUNE
          value: "true"
        - name: EXECUTIONS_DATA_MAX_AGE
          value: "168" # Prune execution logs older than 7 days

Step 2: Separate Webhook Ingestion from UI

By setting EXECUTIONS_PROCESS=main on dedicated webhook ingestion pods, webhook calls return HTTP 200 in sub-10 milliseconds by instantly pushing jobs into Redis, while worker pods scale dynamically based on queue depth.