ZenML Server Deployment Options

ZenML server can be deployed in multiple ways depending on your infrastructure requirements, scale, and operational preferences. This guide covers all deployment options from local development to production-grade setups.

Deployment Methods

Local Server

Quick setup for development and testing

Docker

Containerized deployment for consistency

Kubernetes

Production-grade orchestrated deployment

Cloud Services

Managed deployment on cloud platforms

Local Server Deployment

The simplest way to deploy ZenML server for development and testing.

Quick Start

Deploy a local server with a single command:

zenml up

This command:

Starts a local ZenML server on http://127.0.0.1:8237
Uses SQLite database for storage
Enables local artifact storage
Opens the dashboard in your browser

Custom Configuration

Configure the local server with specific options:

# Deploy on custom port
zenml up --port 8080

# Disable automatic browser opening
zenml up --no-dashboard

# Deploy with specific IP address
zenml up --ip-address 0.0.0.0

# Block until server is stopped
zenml up --blocking

Connecting to Local Server

Connect your ZenML client to the local server:

zenml connect --url http://127.0.0.1:8237

Or use environment variables:

export ZENML_STORE_URL=http://127.0.0.1:8237
export ZENML_STORE_TYPE=rest

Stopping Local Server

Stop the running local server:

zenml down

Docker Deployment

Deploy ZenML server as a Docker container for isolated and reproducible environments.

Using Docker Compose

Create a docker-compose.yml file:

version: '3.8'

services:
  zenml:
    image: zenmldocker/zenml-server:latest
    ports:
      - "8080:8080"
    environment:
      - ZENML_SERVER_AUTH_SCHEME=NO_AUTH
      - ZENML_ANALYTICS_OPT_IN=false
      - ZENML_DEBUG=false
    volumes:
      - zenml-config:/zenml/.zenconfig
      - zenml-data:/zenml/.zenconfig/local_stores

volumes:
  zenml-config:
  zenml-data:

Deploy the server:

docker-compose up -d

Using Docker Run

Run the container directly:

docker run -d \
  --name zenml-server \
  -p 8080:8080 \
  -e ZENML_SERVER_AUTH_SCHEME=OAUTH2_PASSWORD_BEARER \
  -e ZENML_AUTH_JWT_SECRET_KEY=$(openssl rand -hex 32) \
  -v zenml-config:/zenml/.zenconfig \
  -v zenml-data:/zenml/.zenconfig/local_stores \
  zenmldocker/zenml-server:latest

Docker Deployment with MySQL

For production-like Docker deployments, use an external MySQL database:

version: '3.8'

services:
  mysql:
    image: mysql:8.0
    environment:
      - MYSQL_ROOT_PASSWORD=rootpassword
      - MYSQL_DATABASE=zenml
      - MYSQL_USER=zenml
      - MYSQL_PASSWORD=zenmlpassword
    volumes:
      - mysql-data:/var/lib/mysql
    ports:
      - "3306:3306"

  zenml:
    image: zenmldocker/zenml-server:latest
    ports:
      - "8080:8080"
    environment:
      - ZENML_STORE_URL=mysql://zenml:zenmlpassword@mysql:3306/zenml
      - ZENML_SERVER_AUTH_SCHEME=OAUTH2_PASSWORD_BEARER
      - ZENML_AUTH_JWT_SECRET_KEY=${JWT_SECRET_KEY}
      - ZENML_DEFAULT_PROJECT_NAME=default
      - ZENML_ANALYTICS_OPT_IN=true
    depends_on:
      - mysql
    restart: unless-stopped

volumes:
  mysql-data:

Generate JWT secret and start:

export JWT_SECRET_KEY=$(openssl rand -hex 32)
docker-compose up -d

Using ZenML CLI for Docker Deployment

ZenML provides a CLI command to deploy a Docker-based server:

zenml deploy --provider docker

This creates a Docker container with:

SQLite database
Local artifact storage
Exposed on localhost:8238

Accessing Docker Server Logs

View server logs:

# Using docker-compose
docker-compose logs -f zenml

# Using docker
docker logs -f zenml-server

Programmatic Deployment

Deploy ZenML server programmatically using Python:

from zenml.zen_server.deploy.docker import DockerServerDeploymentConfig
from zenml.services import ServiceRegistry

# Configure deployment
config = DockerServerDeploymentConfig(
    name="zenml-server",
    port=8080,
    image="zenmldocker/zenml-server:latest",
)

# Deploy the server
from zenml.zen_server.deploy.docker import DockerZenServer
server = DockerZenServer(config=config)
server.start()

print(f"Server running at: {server.endpoint.config.url}")

# Stop the server when done
server.stop()

Cloud Provider Deployments

AWS Deployment Options

AWS ECS

Deploy on Elastic Container Service

AWS EKS

Deploy on Elastic Kubernetes Service

AWS App Runner

Fully managed container deployment

AWS Fargate

Serverless container deployment

GCP Deployment Options

Cloud Run

Fully managed serverless platform

GKE

Google Kubernetes Engine

Compute Engine

Virtual machine deployment

Cloud SQL

Managed MySQL database

Azure Deployment Options

Container Apps

Fully managed container platform

AKS

Azure Kubernetes Service

Container Instances

Lightweight container deployment

Azure Database

Managed MySQL database

Production Configuration

Essential Environment Variables

Production deployments require proper configuration:

# Authentication
ZENML_SERVER_AUTH_SCHEME=OAUTH2_PASSWORD_BEARER
ZENML_AUTH_JWT_SECRET_KEY=<generate-with-openssl-rand>
ZENML_AUTH_JWT_TOKEN_EXPIRE_MINUTES=60

# Database
ZENML_STORE_URL=mysql://user:pass@host:3306/zenml
ZENML_STORE_POOL_SIZE=20
ZENML_STORE_MAX_OVERFLOW=20
ZENML_STORE_SSL_CA=/path/to/ca.crt

# Server
ZENML_SERVER_URL=https://zenml.example.com
ZENML_SERVER_DEPLOYMENT_TYPE=production
ZENML_DEFAULT_PROJECT_NAME=default

# Secrets Store
ZENML_SECRETS_STORE_TYPE=aws  # or gcp, azure, hashicorp
ZENML_SECRETS_STORE_ENCRYPTION_KEY=<your-encryption-key>

# Performance
ZENML_SERVER_THREAD_POOL_SIZE=40
ZENML_SERVER_AUTH_THREAD_POOL_SIZE=5
ZENML_SERVER_REQUEST_TIMEOUT=20

# Observability
ZENML_ANALYTICS_OPT_IN=true
ZENML_LOGGING_VERBOSITY=INFO

Database Backup Strategy

Configure automatic database backups:

# Backup strategy options:
# - disabled: No backups
# - in-memory: Fast but not persisted
# - dump-file: Persisted to disk/volume
# - database: Copy to backup database
# - mydumper: Using mydumper/myloader tools

ZENML_STORE_BACKUP_STRATEGY=dump-file
ZENML_STORE_BACKUP_DIRECTORY=/backups

SSL/TLS Configuration

Secure database connections with SSL:

# Enable SSL
ZENML_STORE_SSL=true

# SSL certificate paths
ZENML_STORE_SSL_CA=/certs/ca.crt
ZENML_STORE_SSL_CERT=/certs/client-cert.pem
ZENML_STORE_SSL_KEY=/certs/client-key.pem

# Verify server certificate
ZENML_STORE_SSL_VERIFY_SERVER_CERT=true

Resource Requirements

Minimum Requirements (Development)

CPU: 1 core
Memory: 2 GB RAM
Storage: 10 GB
Network: 1 Mbps

Recommended Requirements (Production)

CPU: 4+ cores
Memory: 8+ GB RAM
Storage: 100+ GB
Network: 100+ Mbps
Database: External MySQL with replication

Scaling Guidelines

Users	Pipelines/Day	CPU	Memory	Replicas
1-5	< 50	2	4 GB	1
5-20	50-200	4	8 GB	2
20-50	200-500	8	16 GB	3-5
50+	500+	16+	32+ GB	5-10

Health Checks

Implement health checks for monitoring:

Liveness Probe

curl http://localhost:8080/health

Response:

{
  "status": "ok"
}

Readiness Probe

curl http://localhost:8080/ready

Version Check

curl http://localhost:8080/version

Response:

{
  "version": "0.94.0",
  "deployment": "docker"
}

Troubleshooting

Server Won’t Start

Check logs for errors:

# Local deployment
zenml logs

# Docker deployment
docker logs zenml-server

# Kubernetes deployment
kubectl logs -l app=zenml-server

Database Connection Issues

Verify database connectivity:

# Test MySQL connection
mysql -h host -u user -p -e "SELECT 1;"

# Check ZenML database
mysql -h host -u user -p zenml -e "SHOW TABLES;"

Authentication Failures

Reset admin credentials:

# For local deployments
zenml user create admin --password newpassword

# For containerized deployments
docker exec -it zenml-server zenml user create admin --password newpassword

Port Conflicts

Check if port is already in use:

# Linux/Mac
lsof -i :8080

# Windows
netstat -ano | findstr :8080

Change the port:

zenml up --port 8081

Migration Guide

Migrating from Local to Docker

Export local configuration:

zenml export config --path /tmp/zenml-config

Deploy Docker server:

docker-compose up -d

Import configuration:

docker cp /tmp/zenml-config zenml-server:/tmp/
docker exec zenml-server zenml import config --path /tmp/zenml-config

Migrating from Docker to Kubernetes

Backup database:

docker exec mysql mysqldump -u root -p zenml > zenml-backup.sql

Deploy Kubernetes cluster (see Kubernetes deployment guide)
Restore database:

kubectl exec -i mysql-pod -- mysql -u root -p zenml < zenml-backup.sql

Best Practices

Use External Database

Always use MySQL for production deployments, never SQLite

Enable Authentication

Never deploy without authentication in production environments

Configure SSL/TLS

Use HTTPS and SSL for all network connections

Regular Backups

Implement automated database backup strategies

Monitor Resources

Set up monitoring for CPU, memory, and disk usage

Scale Horizontally

Use multiple replicas for high availability

Rotate Secrets

Regularly rotate JWT keys and database credentials

Update Regularly

Keep ZenML server updated to latest stable version

Next Steps

Kubernetes Deployment

Deploy on Kubernetes with Helm

Docker Guide

Detailed Docker deployment instructions

Security Configuration

Secure your ZenML deployment

Documentation Index

​ZenML Server Deployment Options

​Deployment Methods

Local Server

Docker

Kubernetes

Cloud Services

​Local Server Deployment

​Quick Start

​Custom Configuration

​Connecting to Local Server

​Stopping Local Server

​Docker Deployment

​Using Docker Compose

​Using Docker Run

​Docker Deployment with MySQL

​Using ZenML CLI for Docker Deployment

​Accessing Docker Server Logs

​Programmatic Deployment

​Cloud Provider Deployments

​AWS Deployment Options