System Requirements

This page details all requirements for running LXC AutoScale ML.

Hardware Requirements

Proxmox Host

Resource	Minimum	Recommended
CPU	2 cores	4+ cores
RAM	4 GB	8+ GB
Storage	20 GB free	50+ GB free

The ML model training adds minimal overhead (typically less than 5% CPU during training cycles).

Container Scaling Limits

Default limits that can be adjusted in configuration:

Resource	Minimum	Maximum
CPU Cores	1	8
RAM	512 MB	16 GB

Software Requirements

Proxmox VE

Version	Status
8.x	Tested and supported
7.x	Supported
6.x	Supported
5.x and earlier	Not supported

Python

Python 3.10 to 3.12 is required, and the installation script verifies that Python is available.

The floor comes from the pinned requests, which needs 3.10; the ceiling from the pinned numpy, pandas and scikit-learn, which have no 3.13 wheels. Proxmox VE 8 (Debian 12) ships Python 3.11 and is the primary target. Proxmox VE 9 (Debian 13) ships 3.13, so the ML component needs those three pins raised before it will install there; the API and the monitor are unaffected.

python3 --version

Required Python Packages

These packages are installed automatically by the installation script:

Package	Purpose
flask	API web framework
requests	HTTP client for API calls
scikit-learn	Machine learning (IsolationForest)
pandas	Data manipulation
numpy	Numerical operations
aiofiles	Async file operations
pyyaml	YAML configuration parsing
psutil	System metrics collection
aiohttp	Async HTTP client for batch requests

Optional Packages

Package	Purpose
prometheus-client	Prometheus metrics export

Network Requirements

Inbound Ports

Port	Service	Access
5000	API	Localhost by default

The API ships unauthenticated

server.host defaults to 0.0.0.0 and authentication.enabled defaults to false, so out of the box anything that can route to port 5000 can resize, snapshot, clone and destroy containers. The service runs as root, because it executes pct.

On the standard single-node layout, set server.host: 127.0.0.1 — the model runs beside the API and does not need it exposed. If it must be reachable from another host, enable authentication, set a matching api.api_key in the model configuration, put TLS in front of it, and restrict access with firewall rules.

Outbound Connections

The installation script requires internet access to download packages and source files.

Filesystem Requirements

Required Directories

Directory	Purpose	Permissions
/etc/lxc_autoscale_ml/	Configuration files	root:root 755
/var/log/	Log files	root:root 755
/run/	Lock file (tmpfs, cleared on boot)	root:root 755
/usr/local/bin/lxc_autoscale_ml/	Application code	root:root 755

Disk Space for Logs

File	Typical Size	Notes
lxc_metrics.json	2 MB max	Limited to 1000 entries
lxc_autoscale_ml.log	Varies	Use logrotate
lxc_autoscale_api_access.log	Varies	Use logrotate

Container Requirements

Supported Container Types

• Standard LXC containers
• Privileged and unprivileged containers
• Any Linux distribution inside containers

LXCFS Configuration

For accurate load average reporting inside containers, LXCFS must be configured with the -l flag:

# Check current configuration
grep ExecStart /lib/systemd/system/lxcfs.service

# Should include -l flag:
# ExecStart=/usr/bin/lxcfs /var/lib/lxcfs -l

Verification Commands

Run these commands to verify your system meets requirements:

# Check Proxmox version
pveversion

# Check Python version
python3 --version

# Check available disk space
df -h /var/log

# Check if required commands exist
which pct pvesh

# Verify LXCFS is running
systemctl status lxcfs

Known Limitations

Container ID Range

The API validates container IDs between 100 and 999999. IDs outside this range are rejected.

Maximum Containers

Tested with up to 200 containers. Performance remains stable with batch async processing.

Metrics Storage

The metrics file is limited to 1000 entries (configurable). Older entries are automatically removed.

Next Steps

• Installation: Install LXC AutoScale ML
• Getting Started: Quick start guide