Three-node virtualization clusters are common because they provide a useful quorum model.
But quorum solves only one part of availability.
Consider:
Node 1: 90% RAM
Node 2: 90% RAM
Node 3: 90% RAM
The cluster has quorum.
Then Node 1 fails.
Now the important question is not whether the cluster is quorate.
It is:
Where do Node 1's workloads run?
A useful HA design must account for:
Compute capacity
Can the surviving nodes absorb the protected workloads?
Network
Is cluster communication isolated from congestion?
Storage
Can another host access the VM disks?
Switching
Is there a single upstream switch?
Backup
Can deleted or corrupted data be recovered?
This is why N+1 capacity planning matters.
High availability is not a feature checkbox.
It is a resource and dependency model.
I used a three-node Proxmox architecture as a practical example here:
Top comments (0)