Getting Started

This guide walks you through installing and verifying LXC AutoScale ML on your Proxmox host.

Prerequisites

Before installation, ensure your system meets these requirements:

Requirement	Specification
Proxmox VE	Version 6.x or higher (tested on 8.2.4)
Operating System	Linux (Debian-based)
Python	Version 3.x
Privileges	Root or sudo access
Network	Internet access for package downloads

Installation

Run the installation script on your Proxmox host:

curl -sSL https://raw.githubusercontent.com/fabriziosalmi/proxmox-lxc-autoscale-ml/main/install.sh | bash

The script performs these actions:

1. Verifies system requirements
2. Installs Python dependencies
3. Downloads and configures service files
4. Enables and starts all three services

Dependencies Installed

python3-flask
python3-requests
python3-sklearn
python3-pandas
python3-numpy
python3-aiofiles
python3-yaml
python3-psutil
python3-aiohttp
python3-prometheus-client (optional)

Verify Installation

Check that all services are running:

systemctl status lxc_autoscale_api
systemctl status lxc_monitor
systemctl status lxc_autoscale_ml

Expected output for each service:

● lxc_autoscale_api.service - LXC AutoScale API
     Loaded: loaded
     Active: active (running)

Test the API

Verify the API responds to health checks:

curl http://localhost:5000/health/check

Expected response:

{
  "status": "healthy",
  "timestamp": "2024-12-24T12:00:00Z"
}

Check Metrics Collection

Verify the monitor is collecting metrics:

cat /var/log/lxc_metrics.json | head -20

You should see JSON entries with container metrics.

LXCFS Configuration

Important

Configure LXCFS to report accurate load averages inside containers.

Edit /lib/systemd/system/lxcfs.service and update the ExecStart line to include the -l flag:

[Unit]
Description=FUSE filesystem for LXC
ConditionVirtualization=!container
Before=lxc.service
Documentation=man:lxcfs(1)

[Service]
OOMScoreAdjust=-1000
ExecStartPre=/bin/mkdir -p /var/lib/lxcfs
ExecStart=/usr/bin/lxcfs /var/lib/lxcfs -l
KillMode=process
Restart=on-failure
ExecStopPost=-/bin/fusermount -u /var/lib/lxcfs
Delegate=yes
ExecReload=/bin/kill -USR1 $MAINPID

[Install]
WantedBy=multi-user.target

Apply the changes:

systemctl daemon-reload
systemctl restart lxcfs

Restart your LXC containers for the changes to take effect.

Initial Configuration

API Key Setup

Edit the API configuration to set your API key:

nano /etc/lxc_autoscale_ml/lxc_autoscale_api.yaml

Update the authentication section:

authentication:
  enabled: true
  api_key: "your-secure-api-key-here"

TIP

Generate a secure key using: openssl rand -hex 32

Scaling Thresholds

Review and adjust scaling thresholds in the model configuration:

nano /etc/lxc_autoscale_ml/lxc_autoscale_ml.yaml

Key settings:

scaling:
  cpu_scale_up_threshold: 70      # Scale up at 70% CPU
  cpu_scale_down_threshold: 30    # Scale down at 30% CPU
  ram_scale_up_threshold: 80      # Scale up at 80% RAM
  ram_scale_down_threshold: 40    # Scale down at 40% RAM
  cpu_scale_step: 1               # Add/remove 1 core at a time
  ram_scale_step_mb: 512          # Add/remove 512 MB at a time

Restart services after configuration changes:

systemctl restart lxc_autoscale_api lxc_autoscale_ml

First Scaling Cycle

The system begins collecting metrics immediately. The first scaling decisions occur after sufficient historical data is gathered (typically 5-10 data points).

Monitor the ML service logs to observe scaling activity:

journalctl -u lxc_autoscale_ml -f

Example output:

INFO - Batch fetching configs for 12 containers...
INFO - Batch fetch completed in 0.15s: 12/12 successful
INFO - Processing container 104...
INFO - IsolationForest prediction for 104: -1 (anomaly)
INFO - Scaling decision for 104: CPU=Scale Up (confidence: 87.4%)
INFO - Successfully scaled CPU for LXC ID 104 to 4 cores

Service Management

Start Services

systemctl start lxc_autoscale_api lxc_monitor lxc_autoscale_ml

Stop Services

systemctl stop lxc_autoscale_api lxc_monitor lxc_autoscale_ml

Restart Services

systemctl restart lxc_autoscale_api lxc_monitor lxc_autoscale_ml

View Logs

# API logs
journalctl -u lxc_autoscale_api -n 50

# Monitor logs
journalctl -u lxc_monitor -n 50

# Model logs
journalctl -u lxc_autoscale_ml -n 50

Next Steps

• Architecture: Understand how the components work together
• Configuration: Customize all available settings
• Troubleshooting: Resolve common issues