Production Deployment

Deploy the Predictive Maintenance System to production cloud platforms with automatic scaling, monitoring, and high availability.

Architecture Overview

┌─────────────────┐     ┌─────────────────┐     ┌─────────────────┐
│   Frontend      │     │   Backend       │     │   Database      │
│   (Vercel)      │────▶│   (Render)      │────▶│ (InfluxDB Cloud)│
│   React + Vite  │     │   FastAPI       │     │   Time-Series   │
└─────────────────┘     └─────────────────┘     └─────────────────┘

Component	Technology	Hosting	Status
Frontend	React 18 + Vite	Vercel	✅ Live
Backend	FastAPI + Docker	Render	✅ Live
Database	InfluxDB 2.x	InfluxDB Cloud	✅ Live

Live Deployment

Service	URL
Frontend	https://predictive-maintenance-ten.vercel.app/
Backend API	https://predictive-maintenance-uhlb.onrender.com
API Docs	https://predictive-maintenance-uhlb.onrender.com/docs
Health Check	https://predictive-maintenance-uhlb.onrender.com/health

InfluxDB Cloud Setup

Set up the time-series database for sensor data storage.

Create InfluxDB Cloud account

Go to InfluxDB Cloud
Sign up for a free account
Select AWS us-east-1 region (recommended)

Create a bucket

Navigate to Load Data → Buckets
Click Create Bucket
Name: sensor_data
Retention: 30 days (adjust as needed)

Generate API token

Navigate to Load Data → API Tokens
Click Generate API Token → Read/Write Token
Grant permissions:
- Read access to sensor_data bucket
- Write access to sensor_data bucket
Save the token securely (you won’t see it again)

Note your credentials

Copy these values for later use:

URL: https://us-east-1-1.aws.cloud2.influxdata.com
Organization ID: Found under User → About
Bucket: sensor_data
Token: The token you generated

Backend Deployment (Render)

Deploy the FastAPI backend as a Docker container on Render.

Create Render account

Go to Render
Sign up or log in with GitHub

Create new Web Service

Click New → Web Service
Connect your GitHub repository
Select the repository: PREDICTIVE-MAINTENANCE

Configure service settings

Set the following configuration:

Field	Value
Name	`predictive-maintenance-backend`
Region	Oregon (US West) or closest to you
Branch	`main`
Root Directory	`backend`
Environment	Docker
Dockerfile Path	`backend/Dockerfile`
Instance Type	Free (or paid for production)

Add environment variables

In the Environment section, add:

ENVIRONMENT=production
PORT=8000
INFLUX_URL=https://us-east-1-1.aws.cloud2.influxdata.com
INFLUX_TOKEN=<your-influxdb-token>
INFLUX_ORG=<your-org-id>
INFLUX_BUCKET=sensor_data

Keep INFLUX_TOKEN secure. Never commit it to version control.

Deploy

Click Create Web Service
Render will automatically:
- Build the Docker image from backend/Dockerfile
- Deploy the container
- Assign a URL: https://your-service.onrender.com
Monitor logs for successful startup

Verify deployment

Test the health endpoint:

curl https://your-service.onrender.com/health

Expected response:

{
  "status": "healthy",
  "db_connected": true
}

Render Cold Starts

Render free tier services spin down after 15 minutes of inactivity. First request after idle may take 30-60 seconds.

The frontend automatically sends a 10-minute keep-alive ping to /ping to prevent cold starts during active browser sessions.

Frontend Deployment (Vercel)

Deploy the React dashboard to Vercel with automatic HTTPS and global CDN.

Create Vercel account

Go to Vercel
Sign up with GitHub

Import repository

Click Add New → Project
Import your GitHub repository: PREDICTIVE-MAINTENANCE

Configure build settings

Set the following configuration:

Field	Value
Framework Preset	Vite
Root Directory	`frontend`
Build Command	`npm run build`
Output Directory	`dist`
Install Command	`npm install`

Add environment variables (optional)

In Settings → Environment Variables, add:

VITE_API_URL=https://your-backend.onrender.com

This is optional if you’re using Vercel rewrites (recommended approach).

Configure API rewrites

The frontend/vercel.json file handles API proxying:

{
  "rewrites": [
    {
      "source": "/api/:path*",
      "destination": "https://predictive-maintenance-uhlb.onrender.com/api/:path*"
    },
    {
      "source": "/system/:path*",
      "destination": "https://predictive-maintenance-uhlb.onrender.com/system/:path*"
    },
    {
      "source": "/health",
      "destination": "https://predictive-maintenance-uhlb.onrender.com/health"
    },
    {
      "source": "/ping",
      "destination": "https://predictive-maintenance-uhlb.onrender.com/ping"
    }
  ]
}

Update the destination URLs to match your Render backend URL.

Deploy

Click Deploy
Vercel will:
- Install dependencies
- Build the React app
- Deploy to global CDN
- Assign a URL: https://your-project.vercel.app
Automatic deployments trigger on every push to main

Test the deployment

Open your Vercel URL
Verify the dashboard loads
Check browser DevTools Network tab
Ensure API calls proxy to Render backend

Custom Domain (Optional)

Add domain in Vercel

Go to Settings → Domains
Add your custom domain (e.g., pm.yourcompany.com)

Update DNS

Add these DNS records at your domain provider:

Type	Name	Value
A	pm	76.76.21.21
CNAME	pm	cname.vercel-dns.com

Wait for verification

Vercel will automatically provision SSL certificates via Let’s Encrypt.

Systemd Deployment (Linux Servers)

For bare-metal Linux servers, use the systemd service for process management.

Run setup script

The automated setup script installs dependencies and configures systemd:

sudo ./scripts/setup_linux.sh

This script:

Installs Python 3.11+, pip, venv
Creates /opt/predictive-maintenance directory
Sets up Python virtual environment
Installs backend dependencies
Configures systemd service
Starts the service automatically

Verify service status

sudo systemctl status predictive-maintenance

Expected output:

● predictive-maintenance.service - Predictive Maintenance Backend API
   Loaded: loaded (/etc/systemd/system/predictive-maintenance.service)
   Active: active (running) since Mon 2026-03-02 12:00:00 UTC

Manage the service

sudo systemctl start predictive-maintenance

Systemd Service Configuration

The service is defined in /etc/systemd/system/predictive-maintenance.service:

[Unit]
Description=Predictive Maintenance Backend API
After=network.target influxdb.service
Wants=influxdb.service

[Service]
Type=simple
User=root
Group=root
WorkingDirectory=/opt/predictive-maintenance
Environment="PATH=/opt/predictive-maintenance/venv/bin"
Environment="PYTHONPATH=/opt/predictive-maintenance"
ExecStart=/opt/predictive-maintenance/venv/bin/uvicorn backend.api.main:app --host 0.0.0.0 --port 8000
Restart=always
RestartSec=5
StandardOutput=journal
StandardError=journal

[Install]
WantedBy=multi-user.target

Key features:

Restart=always - Auto-restart on failure
RestartSec=5 - Wait 5 seconds before restart
Logs to systemd journal
Starts after network is available

Environment Variables Reference

Backend (Render)

Variable	Required	Description	Example
`ENVIRONMENT`	Yes	Runtime mode	`production`
`PORT`	Yes	Server port	`8000`
`INFLUX_URL`	Yes	InfluxDB Cloud URL	`https://us-east-1-1.aws.cloud2.influxdata.com`
`INFLUX_TOKEN`	Yes	InfluxDB API token	`kg2i8Mq...`
`INFLUX_ORG`	Yes	Organization ID	`67c4314d97304c09`
`INFLUX_BUCKET`	Yes	Bucket name	`sensor_data`

Frontend (Vercel)

Variable	Required	Description	Example
`VITE_API_URL`	No	Backend URL (if not using rewrites)	`https://your-backend.onrender.com`

Monitoring and Observability

Health Checks

curl https://your-backend.onrender.com/health

Logs

Render: Access logs via dashboard → Service → Logs Vercel: View logs in dashboard → Project → Functions InfluxDB: Monitor queries at https://cloud2.influxdata.com Systemd: View logs with:

sudo journalctl -u predictive-maintenance -f --since "1 hour ago"

Metrics

InfluxDB Cloud provides built-in metrics:

Data points written per second
Query response time
Storage usage
API request rate

Troubleshooting

Backend 503 Errors

Problem: Service unavailable on Render Solution: Check deployment logs for:

InfluxDB connection failures
Missing environment variables
ML model loading errors (lazy imports)

curl https://your-backend.onrender.com/health -v

Frontend API Errors

Problem: Network Error or CORS issues Solution: Verify Vercel rewrites are working:

Check frontend/vercel.json destination URLs
Inspect browser DevTools → Network tab
Ensure requests to /api/* proxy to Render

InfluxDB Token Scope Issues

Problem: unauthorized access errors Solution: Regenerate token with correct permissions:

Go to InfluxDB Cloud → API Tokens
Create new Read/Write Token
Grant access to sensor_data bucket
Update INFLUX_TOKEN in Render environment variables
Restart the service

Degradation Index (DI) State Loss

Problem: Health score resets after backend restart Solution: The system automatically hydrates DI from InfluxDB on startup via |> last(). If this fails:

Check InfluxDB connection
Verify degradation_index measurement exists
Use POST /system/purge to explicitly reset DI to 0.0

Windows Binary Errors on Vercel

Problem: Error 126 during build Solution: Remove node_modules from Git:

git rm -r --cached frontend/node_modules
echo "node_modules/" >> frontend/.gitignore
git commit -m "fix: remove node_modules from tracking"
git push

Redeploy on Vercel.

Performance Optimization

Backend

Enable Gunicorn for production (multi-worker):

gunicorn backend.api.main:app -w 4 -k uvicorn.workers.UvicornWorker

Cache ML models in memory (already implemented)
Batch InfluxDB writes for high-frequency data
Use connection pooling for database clients

Frontend

Enable Vercel Edge Caching for static assets
Lazy load components using React.lazy()
Optimize chart rendering with data sampling
Use service workers for offline capability

Database

Downsample old data using InfluxDB tasks:

from(bucket: "sensor_data")
  |> range(start: -30d)
  |> aggregateWindow(every: 1h, fn: mean)
  |> to(bucket: "sensor_data_downsampled")

Set appropriate retention policies (30 days for raw data)
Use continuous queries for pre-aggregation

Security Best Practices

Always follow these security guidelines in production:

Never commit secrets to version control
- Use .env files locally
- Use platform environment variables in production
Rotate API tokens regularly
- InfluxDB tokens every 90 days
- Backend service credentials quarterly

Enable CORS restrictions in backend/api/main.py:

CORS_ORIGINS = [
    "https://your-frontend.vercel.app",
]

Use HTTPS everywhere
- Vercel provides automatic SSL
- Render provides automatic SSL
- InfluxDB Cloud enforces HTTPS
Implement rate limiting using FastAPI middleware
Validate all inputs with Pydantic schemas
Monitor for anomalous API usage in logs

Scaling Considerations

Horizontal Scaling

Render: Upgrade to paid plan for:

Auto-scaling based on CPU/memory
Multiple instances with load balancing
99.9% uptime SLA

Vercel: Automatic edge network scaling InfluxDB Cloud: Auto-scales with usage-based pricing

Vertical Scaling

For high-frequency data (>1000 data points/sec):

Upgrade Render instance type (2GB → 4GB RAM)
Use InfluxDB Dedicated instances
Implement data batching in backend

Next Steps

Review API Reference for integration
Explore Docker Deployment for local testing
Check Architecture for system design
Monitor your deployment with the built-in health dashboard

Get Started

Deployment

Core Features

Machine Learning

Dashboard

Architecture Overview

Live Deployment

InfluxDB Cloud Setup

Backend Deployment (Render)

Render Cold Starts

Frontend Deployment (Vercel)

Custom Domain (Optional)

Systemd Deployment (Linux Servers)

Systemd Service Configuration

Environment Variables Reference

Backend (Render)

Frontend (Vercel)

Monitoring and Observability

Health Checks

Logs

Metrics

Troubleshooting

Backend 503 Errors

Frontend API Errors

InfluxDB Token Scope Issues

Degradation Index (DI) State Loss

Windows Binary Errors on Vercel

Performance Optimization

Backend

Frontend

Database

Security Best Practices

Scaling Considerations

Horizontal Scaling

Vertical Scaling

Next Steps

Build docs developers (and LLMs) love

Get Started

Deployment

Core Features

Machine Learning

Dashboard

​Architecture Overview

​Live Deployment

​InfluxDB Cloud Setup

​Backend Deployment (Render)

​Render Cold Starts

​Frontend Deployment (Vercel)

​Custom Domain (Optional)

​Systemd Deployment (Linux Servers)

​Systemd Service Configuration

​Environment Variables Reference

​Backend (Render)

​Frontend (Vercel)

​Monitoring and Observability

​Health Checks

​Logs

​Metrics

​Troubleshooting

​Backend 503 Errors

​Frontend API Errors

​InfluxDB Token Scope Issues

​Degradation Index (DI) State Loss

​Windows Binary Errors on Vercel

​Performance Optimization

​Backend

​Frontend

​Database

​Security Best Practices

​Scaling Considerations

​Horizontal Scaling

​Vertical Scaling

​Next Steps

Build docs developers (and LLMs) love

Architecture Overview

Live Deployment

InfluxDB Cloud Setup

Backend Deployment (Render)

Render Cold Starts

Frontend Deployment (Vercel)

Custom Domain (Optional)

Systemd Deployment (Linux Servers)

Systemd Service Configuration

Environment Variables Reference

Backend (Render)

Frontend (Vercel)

Monitoring and Observability

Health Checks

Logs

Metrics

Troubleshooting

Backend 503 Errors

Frontend API Errors

InfluxDB Token Scope Issues

Degradation Index (DI) State Loss

Windows Binary Errors on Vercel

Performance Optimization

Backend

Frontend

Database

Security Best Practices

Scaling Considerations

Horizontal Scaling

Vertical Scaling

Next Steps