Deploying Velda¶

This section covers deploying and managing self-hosted Velda clusters.

Deployment Options¶

Velda supports several deployment methods:

1. Terraform Deployment¶

Infrastructure-as-code deployment: - Automated deployment with major cloud providers - Reproducible & updatable infrastructure configuration - Suitable for production workloads

Check out Terraform templates to get started.

2. Marketplace (Coming soon)¶

We're working with most cloud providers to provide marketplace options for quick setup.

3. Velda Cloud¶

Managed service option that is ready to run: - Hosted infrastructure - GPU and CPU resources available - Includes free tier with monthly credits - Suitable for individuals and small teams, or exploration.

Get Started with Velda Cloud

4. Manual Cluster Setup (Custom Deployments)¶

Direct installation on Linux machines: - Install on bare metal or VMs - Full control over all components - Customizable configuration - Suitable for on-premises deployments

Manual Setup Guide

Architecture Overview¶

A Velda cluster consists of:

API Server: Control plane that manages scheduling, authentication, and orchestration
Agents: Worker nodes that execute containers and report resource status
Storage: ZFS-based storage shared via NFS for user data and instances. Typically attached with controller. -- SSH client: Use a standard SSH client to connect to cluster nodes

                         ┌────────────────────────────────────┐
                         │            Velda Cluster           │
                         │                                    │
                         │  ┌─────────────┐   ┌─────────────┐  │
                         │  │ API Server  │◄──┤ Agent Pool  │  │
                         │  │  (Control)  │   │  (Workers)  │  │
                         │  └─────────────┘   └─────────────┘  │
                         │         │           ▲                │
                         │         │           │                │
                         │  ┌─────────────┐    │                │
                         │  │  Storage    │    │                │
                         │  │ (ZFS + NFS) │    │                │
                         │  └─────────────┘    │                │
                         └────────────────────────────────────┘
                                   ▲
                                   │ SSH
                            ┌─────────────┐
                            │ SSH client  │
                            └─────────────┘

Choosing a Deployment Method¶

Deployment Method	Use Case	Setup Time	Scalability	Management
Velda Cloud	Getting started, small teams	< 5 min	Auto-scaling	Managed
Terraform	Production, repeatable infra	15-30 min	Highly scalable	IaC managed
Manual Setup	Custom deployments, on-prem	30-60 min	Configurable	Self-managed

System Requirements¶

For Self-Hosted Deployments¶

API Server (Control Plane): - Linux (Ubuntu 22.04+ recommended) - 2+ vCPUs, 4+ GB RAM - Dedicated disk for ZFS storage - Network accessible by all agents and clients

Agent Nodes (Workers): - Linux (Ubuntu 22.04+ recommended) - AMD64 or ARM64 architecture - 1+ vCPUs, 4+ GB RAM (more for GPU workloads) - CGroup v2 enabled (CGroup v1 disabled) - Direct IP access to API server

For GPU Workloads: - NVIDIA GPU drivers