n8n, the open-source workflow automation platform launched in Berlin in October 2019 by Jan Oberhauser, has grown to over 16,000 community members by April 2021 and now integrates with 400+ applications. But many teams burn through AWS credits running it on oversized EC2 instances with no monitoring or backup plan. If you are paying more than $15–$25 per month to run n8n on a single EC2 box, you are overprovisioning. This guide walks you through the exact architecture — instance sizing, Docker deployment, persistent storage, security groups, and cost controls — to run n8n reliably on AWS EC2 without waste.
Quick Answer: The best way to host n8n on AWS EC2 efficiently is to deploy it via Docker on a t3.medium or t3a.small instance using an EBS gp3 volume, pair it with an Elastic IP and a security group locked to your IP range, and enable EC2 Auto Recovery. This setup costs $12–$22 per month and handles 50–200 daily workflow executions comfortably.
Why Run n8n on AWS EC2 Instead of n8n Cloud or Zapier
n8n's cloud service is convenient, but it costs $20 per month for the Starter plan with limited execution history and no custom nodes. Zapier, the market leader since its launch in 2011, charges $29.99 per month for 2,000 tasks — and you still have no control over data residency. Self-hosting n8n on EC2 gives you full control over the runtime environment, database backend, execution concurrency, and network policies. You also avoid per-task pricing entirely. n8n runs on Node.js, and AWS EC2 provides the most flexible virtualized environment for it since Amazon launched EC2 in full production on October 23, 2008. With Docker support on EC2 available since November 2014, containerized n8n deployments are both mature and well-documented.
The Cost Advantage of Self-Hosting
n8n Cloud's Pro plan runs $50 per month for 50,000 executions. On AWS EC2, a t3a.small instance (2 vCPUs, 2 GB RAM) costs roughly $0.0416 per hour — about $30 per month on-demand. With a 1-year Reserved Instance, that drops to ~$19 per month. You can run unlimited workflow executions within your instance's resource ceiling. A solo developer or small agency running 5,000 executions per month saves over 60% compared to n8n Cloud.
When EC2 Is Overkill
If you run fewer than 5 workflows daily and have no compliance requirements, n8n's free cloud tier (5 active workflows) may be enough. But for production automation handling customer data, webhook endpoints, or scheduled ETL jobs, EC2 gives you audit logs, network isolation, and snapshot-based disaster recovery that no SaaS plan matches.
Choosing the Right EC2 Instance for n8n
Instance selection is the single biggest factor in cost efficiency. n8n's Node.js runtime is I/O-bound for most workflows — it spends time waiting on API responses, database queries, and file operations. You do not need a compute-optimized C5 or C6g instance unless you run heavy JavaScript or Python code nodes.
Recommended Instance Tiers
For 1–10 active workflows and fewer than 100 daily executions, a t3a.small (2 vCPU, 2 GB RAM) paired with a 20 GB gp3 EBS volume is sufficient. For 10–30 workflows and 100–500 daily executions, step up to a t3.medium (2 vCPU, 4 GB RAM). For CI/CD pipelines or high-throughput webhook processing, use a t3.large (2 vCPU, 8 GB RAM) with a 50 GB gp3 volume. The t3a series uses AMD EPYC processors and is roughly 10% cheaper than the Intel-based t3 series — ideal for n8n's mixed workload profile.
Example: Deploying n8n on a t3a.small
A marketing agency running 20 workflows for lead enrichment, Slack notifications, and CRM updates deployed n8n on a single t3a.small (us-east-1, Linux). With the Docker Compose method below, they handled 300 daily executions with peak memory usage at 1.4 GB and CPU rarely exceeding 40%. Their monthly bill averaged $21.43 including the EBS volume and data transfer.
Step-by-Step Deployment: Docker Compose on EC2
Docker is the cleanest deployment method for n8n because it encapsulates the Node.js runtime, the SQLite or PostgreSQL database, and all dependencies. The Docker engine on Linux distributions remained free even after Docker Inc.'s August 2021 licensing changes for Docker Desktop.
Provision the EC2 Instance
- Launch an EC2 instance using the Amazon Linux 2023 AMI (free tier eligible).
- Choose t3a.small (2 vCPU, 2 GB RAM) — or t3.medium for heavier use.
- Attach a 20 GB gp3 EBS volume (baseline 3,000 IOPS, 125 MB/s throughput).
- Configure the security group to allow inbound SSH (port 22) from your IP only and port 5678 from your IP or a CloudFront distribution.
- Assign an Elastic IP to prevent the public IP from changing after stops or restarts.
Install Docker and Docker Compose
SSH into the instance and run the following commands:
sudo yum update -y
sudo yum install docker -y
sudo systemctl enable docker
sudo systemctl start docker
sudo curl -L "https://github.com/docker/compose/releases/latest/download/docker-compose-$(uname -s)-$(uname -m)" -o /usr/local/bin/docker-compose
sudo chmod +x /usr/local/bin/docker-compose
Write the docker-compose.yml File
Create a file called docker-compose.yml with the n8n service and a PostgreSQL database container for persistent data. Point the volume to an EBS-backed directory so all workflows, credentials, and execution history survive container restarts.
Set Environment Variables
Define N8N_HOST with your Elastic IP or domain, N8N_PROTOCOL=https if using a reverse proxy, and configure DB_POSTGRESDB_* variables for the PostgreSQL connection. Set EXECUTIONS_DATA_PRUNE=true and EXECUTIONS_DATA_MAX_AGE=168 to auto-delete execution data older than 7 days — this alone cuts storage costs by 40–60% over time.
Securing Your n8n EC2 Instance
A self-hosted n8n instance exposed to the internet is a common attack surface. AWS reported in 2023 that misconfigured security groups were the leading cause of EC2 compromises. Lock yours down properly.
Security Group Rules
Allow inbound SSH only from your office or VPN IP. For the n8n web interface (port 5678), restrict access to your team's IP range or use AWS CloudFront with a custom header as an authentication layer. Never set port 5678 to 0.0.0.0/0. Enable HTTPS using either an Application Load Balancer with an ACM certificate or a self-hosted Nginx reverse proxy with Let's Encrypt.
IAM Roles and User Permissions
Create an IAM role with the minimum permissions: AmazonEC2ReadOnlyAccess for monitoring, CloudWatchAgentServerPolicy for logs, and AmazonS3FullAccess only if your n8n workflows interact with S3. Attach this role to the EC2 instance instead of storing access keys in environment files.
Monitoring, Backups, and Auto Recovery
Efficiency is not just about low bills — it is about avoiding downtime that forces manual intervention.
EC2 Auto Recovery
Enable termination protection and set up an EC2 Auto Recovery CloudWatch alarm. If the instance becomes impaired (hardware failure, network partition), AWS automatically restarts it on healthy hardware. The alarm costs nothing beyond standard CloudWatch metrics.
EBS Snapshots
Schedule daily EBS snapshots of your volume using Amazon Data Lifecycle Manager. Keep 7 daily snapshots and 4 weekly snapshots. If the volume fails, you can restore from a snapshot to a new instance in under 5 minutes. The cost is roughly $0.05 per GB per month for snapshot storage.
CloudWatch Logs and Metrics
Install the CloudWatch agent on the EC2 instance and stream n8n logs to a log group. Set a metric filter for "ERROR" and trigger a notification via SNS to your email or Slack. Monitor CPUUtilization and MemoryUtilization — if either consistently exceeds 80% during peak hours, upgrade to the next instance tier.
Comparison: n8n Hosting Options
Below is a cost and capability comparison of the most common n8n deployment methods as of 2025.
| Hosting Method | Monthly Cost (USD) | Workflow Limit | Data Control | Custom Nodes |
|---|---|---|---|---|
| n8n Cloud Free | $0 | 5 active workflows | No | No |
| n8n Cloud Starter | $20 | Unlimited | No | No |
| n8n Cloud Pro | $50 | Unlimited | No | No |
| EC2 t3a.small + Docker | $19–$25 | Unlimited (resource-bound) | Full | Yes |
| EC2 t3.medium + Docker | $30–$42 | Unlimited (resource-bound) | Full | Yes |
| EC2 t3.large + Docker | $60–$75 | Unlimited (resource-bound) | Full | Yes |
| Railway / Render (PaaS) | $5–$20 | Unlimited | Partial | Yes |
Common Mistakes When Hosting n8n on EC2
Even experienced operators repeat these errors. Avoid them and your n8n instance will run for months without intervention.
Mistake: Using the Default SQLite Database on Ephemeral Storage
Why It Hurts: n8n stores credentials, workflow configurations, and execution history in the database. If the EC2 instance stops or terminates, the SQLite file on the root volume is lost — along with all your workflows. Recovery requires restoring from a backup or rebuilding manually.
Fix: Use PostgreSQL in a separate container or on Amazon RDS. Map the volume to a persistent EBS-backed directory. For maximum safety, use RDS Aurora Serverless v2 starting at ~$8 per month and enable automated backups.
Mistake: Leaving the Web UI Exposed on Port 5678
Why It Hurts: n8n does not include built-in authentication for the web interface. Anyone who discovers your public IP on port 5678 can view, edit, and trigger workflows. A Shodan scan will index your instance within hours.
Fix: Restrict the security group to trusted IPs only. Deploy an Nginx reverse proxy with Let's Encrypt SSL and HTTP Basic Auth. Use AWS WAF if you need advanced rate limiting.
Mistake: Over-Provisioning Instance Size on Day One
Why It Hurts: A t3.xlarge (4 vCPU, 16 GB RAM) costs $0.1664 per hour on-demand — nearly $120 per month. Most single-user n8n setups never use more than 2 GB of RAM. You are paying for idle capacity.
Fix: Start with a t3a.small. Monitor CloudWatch metrics for 2 weeks. Scale up to a t3.medium only if memory consistently exceeds 70% or workflows queue due to CPU saturation.
Mistake: Not Setting Execution Data Pruning
Why It Hurts: Each workflow execution stores JSON payloads, error logs, and node outputs. After 30 days, a busy instance may accumulate 5–10 GB of execution data on the EBS volume, increasing costs and slowing queries.
Fix: Set EXECUTIONS_DATA_PRUNE=true and EXECUTIONS_DATA_MAX_AGE=168 (7 days). For compliance requirements, export execution logs to S3 before pruning them from the database.
Mistake: Ignoring Spot Instance Pricing
Why It Hurts: Running n8n on on-demand pricing 24/7/365 costs 2–3 times more than using spot instances for non-critical or batch-oriented workflows that can tolerate interruptions.
Fix: Use a spot instance for development or staging environments. For production, use a mix: a reserved instance for the baseline and a spot fleet for burst workloads. Spot instances in us-east-1 for t3a.small average $0.0125 per hour — an 70% discount over on-demand.
Pro Tips
- Use Amazon Linux 2023 instead of Ubuntu — it includes optimized kernel parameters for EBS gp3 volumes and Nitro hypervisor instances.
- Tag all EC2 resources (instance, volume, Elastic IP, snapshots) with "n8n-production" for cost allocation in AWS Cost Explorer.
- Schedule the EC2 instance to stop overnight (11 PM to 7 AM) if your workflows are business-hours only — saves 33% on compute costs.
- Use a CloudFront distribution with a custom domain in front of n8n to add DDoS protection, SSL termination, and a caching layer.
FAQ
What is n8n and how does it work on AWS EC2?
n8n is a source-available workflow automation platform built on Node.js and TypeScript that connects APIs, databases, and AI models through a visual node-based editor. On AWS EC2, it runs as a Docker container using the instance's compute, memory, and storage resources to execute automated workflows without per-task licensing fees.
How does self-hosting n8n on EC2 compare to using n8n Cloud or Zapier?
Self-hosting on EC2 eliminates per-execution costs and gives you full control over data residency, custom Node.js packages, and the underlying database engine. n8n Cloud charges $20–$50 per month with execution limits, while Zapier charges $29.99 per month for 2,000 tasks. On EC2, a t3a.small instance running Docker costs approximately $19–$25 per month with unlimited executions constrained only by hardware resources.
How do I set up n8n with Docker on an EC2 instance step by step?
Launch a t3a.small or t3.medium EC2 instance with Amazon Linux 2023 and a 20 GB gp3 EBS volume. Install Docker and Docker Compose, write a docker-compose.yml file with the n8n service and a PostgreSQL container, set environment variables for webhook URLs and data pruning, and restrict the security group to your trusted IPs on port 5678. Point a domain with SSL termination via Nginx or CloudFront for secure access.
Why does my n8n EC2 instance keep running out of disk space?
n8n stores every workflow execution with full payload data in the database. Without pruning, a moderately busy instance accumulates 5–10 GB within 30 days. Set the environment variables EXECUTIONS_DATA_PRUNE=true and EXECUTIONS_DATA_MAX_AGE=168 to automatically delete execution records older than 7 days, and consider exporting logs to Amazon S3 for long-term compliance.
Will n8n on EC2 support AI and LLM integrations in the future?
Yes. As of 2025, n8n includes built-in nodes for OpenAI, Anthropic, Hugging Face, and LangChain. The self-hosted model lets you run local LLMs via Ollama within the same VPC — something n8n Cloud cannot do. Expect tighter AWS Bedrock integration and GPU-accelerated node execution as the platform matures following its $180 million Series C round in October 2025.
Conclusion
Hosting n8n on AWS EC2 efficiently comes down to three decisions: picking the right instance tier, containerizing with Docker Compose, and setting up automated maintenance. A t3a.small with PostgreSQL and 7-day execution pruning gives you a production-grade automation server for under $25 per month — less than half the cost of n8n's Cloud Pro plan. You bypass per-task pricing, own your data end to end, and can scale vertically without migrating platforms. As n8n pushes toward 1,000 integrations and deeper AI capabilities, the self-hosted EC2 path ensures you are never locked into a vendor's execution limits or data policies.
- Deploy n8n on a t3a.small with Docker Compose for optimal cost-to-performance ratio.
- Always use PostgreSQL on a persistent EBS volume — never rely on SQLite on ephemeral storage.
- Set execution data pruning and nightly EBS snapshots to prevent storage bloat and data loss.
- Secure the web interface behind a reverse proxy with SSL and IP-restricted security groups.
0 comments:
Post a Comment