C P M

Building a Proxmox VE High-Availability Cluster: Enterprise Setup Guide

Blog 8 - CodeProMax Tech

Resilient Virtualization with Proxmox VE Clusters

Direct Answer: A Proxmox VE High-Availability (HA) cluster provides automated virtual machine failover by linking three or more nodes via Corosync clustering software and redundant low-latency network fabrics, backed by distributed Ceph or ZFS shared storage.

1. Quorum Architecture & The Rule of Three

Proxmox uses Corosync for cluster membership and quorum voting. In a 2-node cluster, a single node loss leaves the surviving node with only 50% of votes, leading to “split-brain” where the surviving node refuses to start VMs to prevent storage corruption. For true high availability, an odd number of nodes (minimum 3) or an external low-power QDevice witness is mandatory.

Component Minimum Hardware Enterprise Production Spec
Nodes 3 Physical Nodes 3 to 7 Homogeneous Enterprise Rack Servers
Corosync Network 1GbE Dedicated Ethernet Dual Low-Latency Ring (Ring 0 & Ring 1) on Dedicated VLAN
Storage Subsystem Shared NFS / iSCSI SAN Ceph HCI (3x OSDs per node on Enterprise NVMe SSDs)
Fencing / Watchdog Software Softdog Hardware IPMI / BMC Watchdog for instantaneous STONITH

2. Setting Up Corosync Redundant Network Rings

# Create cluster on Node 1 with dual corosync links
pvecm create prod-cluster --link0 10.10.10.11 --link1 10.10.20.11

# Add Node 2 with link redundancy
pvecm add 10.10.10.11 --link0 10.10.10.12 --link1 10.10.20.12

# Add Node 3
pvecm add 10.10.10.11 --link0 10.10.10.13 --link1 10.10.20.13

# Verify cluster quorum status
pvecm status

3. Ceph Storage Integration & VM Failover Times

With distributed Ceph storage, VM virtual disks are replicated synchronously across all three nodes. If Node 1 experiences a sudden motherboard or power failure:

  1. The hardware watchdog detects missing heartbeats within 60 seconds.
  2. The cluster manager fences the failed node (STONITH).
  3. Nodes 2 and 3 retain quorum (66.6% majority) and automatically restart the failed VMs from the latest Ceph replicated state. Total downtime is under 90 seconds.

Need Production-Grade Home & Production Server Infrastructure?

At CodeProMax Tech, our senior infrastructure and software engineers design, implement, and maintain tailored enterprise solutions with 24/7 SLA engineering support.

Explore Home & Production Server Infrastructure →

CodeProMax Tech Technical Author
Published By
Admin
Leave a Comment

Talk with an ERP & DevOps Specialist