Skip to main content

Multi-Node Configuration

Clustron DKV allows you to run stores across multiple nodes, either on a single machine or across multiple machines.

Understanding how to configure nodes correctly is key to building a scalable system.

Two Deployment Models​

Clustron supports two common setups:

1. Single Machine (Multiple Nodes)​

You can run multiple nodes on the same machine.

  • One manager per machine
  • Multiple nodes controlled using InstanceCount
Machine
  ├── Manager (7801)
  ├── node1 (7811 / 7861)
  └── node2 (7812 / 7862)

How to Configure​

New-DkvStore `
  -Name TestStore `
  -InstancePrefix node `
  -InstanceCount 2 `
  -ClustronPort 7811 `
  -ClientPort 7861

When to Use​

  • Local development
  • Testing distributed behavior
  • Running multiple nodes without additional machines

2. Multiple Machines (Servers)​

You can run nodes across multiple machines.

  • One manager per machine
  • Typically one node per machine
Machine A → Manager + node1  
Machine B → Manager + node2

How to Configure​

Run on each machine:

New-DkvStore `
  -Name TestStore `
  -InstancePrefix node `
  -InstanceCount 1 `
  -ClustronPort 7811 `
  -ClientPort 7861

Manager Connection​

You must connect to all managers:

Connect-DkvManager -Managers 10.0.0.11:7801,10.0.0.12:7801

When to Use​

  • Production deployments
  • High availability
  • Scaling across servers

How Nodes Are Identified​

Nodes are named using the prefix:

InstancePrefix = node

node1
node2
node3

Port Assignment​

Ports are assigned per node using incrementing values.

ClustronPort = 7811 → 7811, 7812, ...
ClientPort   = 7861 → 7861, 7862, ...

Each node gets unique ports automatically.

Choosing the Right Setup​

ScenarioRecommended Setup
Local testingSingle machine (multi-node)
Small deployment1 node per machine
High scalabilityMultiple machines
Fault tolerance requiredMultiple machines

Key Principles​

  • One manager per machine
  • Nodes are defined using InstanceCount
  • Ports must not overlap
  • Clients can connect to any node

Common Mistakes​

Using Multiple Managers on One Machine​

  • Only one manager should run per machine

Incorrect InstanceCount​

  • Too high → port conflicts
  • Too low → insufficient capacity

Not Connecting All Managers​

  • Required in multi-machine setups
  • Ensures proper cluster coordination

What’s Next​

👉 Continue to Ports & Networking to understand how communication works