Architecture

Overview

This repository provides two primary architectures for deploying Ignition on Kubernetes.

1. Failover (Master/Backup)

The ignition-failover chart deploys a standard redundant pair of Ignition gateways.

• Replicas: 2 (Fixed)

• State: StatefulSet with stable network identities.

• Networking:

  • 0: Master (Primary)
  • 1: Backup (Redundant)

• Direct Access: In addition to the main load-balanced Service, the chart creates two Headless Services for direct pod targeting:

  • {{release-name}}-primary: Always points to ordinal 0.
  • {{release-name}}-backup: Always points to ordinal 1.

• Use Case: Standard SCADA deployments requiring high availability.

graph TD
    User((User/Device)) --> Ingress
    Ingress --> Service
    subgraph K8s Cluster
        Service --> Pod0[Ignition-0: Master ]
        Service -.-> Pod1[Ignition-1: Backup ]
        Pod0 <-->|Gateway Network| Pod1
    end

2. Scaleout (Frontend/Backend)

The ignition-scaleout chart separates the workload into two distinct StatefulSets:

• Backend: Handles device connections, database logging, and tag history.
• Frontend: Handles Perspective sessions and API requests.

This architecture allows you to scale the Frontend layer independently of the Backend layer to handle massive user loads.

graph TD
    User((User)) --> LB[Load Balancer]
    subgraph Frontend Layer
        LB --> FE1[Frontend-0]
        LB --> FE2[Frontend-1]
        LB --> FE3[Frontend-2]
    end
    subgraph Backend Layer
        FE1 -->|GAN| BE[Backend: Master/Backup]
        FE2 -->|GAN| BE
        FE3 -->|GAN| BE
        Device((PLC/Device)) --> BE
    end

Selection Guide

Choosing the right architecture depends on your specific requirements for scale, latency, and management complexity.

Decision Matrix

Feature	Failover (Master/Backup)	Scaleout (Frontend/Backend)
Primary Goal	High Availability (HA)	High Concurrency (User Load)
Max User Load	~500 Concurrent Sessions	1,000+ Concurrent Sessions
Complexity	Low	High
Device Load	High	High (Isolated to Backend)
Maintenance	Simple (2 nodes)	Complex (N+2 nodes)
Licencing	1 Redundant Pair	1 Backend Pair + N Frontend Licences

When to choose Failover

• You are deploying a standard factory-floor SCADA system.
• Your user count is moderate (under 500 concurrent).
• Simplicity of maintenance and licencing is a priority.

When to choose Scaleout

• You are building an enterprise-wide dashboarding solution.
• You expect thousands of users to access Perspective sessions.
• You need to protect critical device communications (Backend) from heavy user query loads (Frontend).