Universal Cluster Architecture¶

What Are Clusters?¶

Clusters are a DRP coordination mechanism for managing groups of machines as a unit. A Cluster object has its own workflow and lifecycle that operates independently from the individual machines it contains.

This enables capabilities such as: - Building, resizing, or destroying multi-machine deployments - Managing platform APIs (Kubernetes, OpenShift) at the cluster level - Performing cluster-level operations without direct machine interaction

How Clusters Work¶

Each machine in a cluster references the cluster's tracking profile via the cluster/tags parameter. The Cluster object orchestrates cross-machine operations through its own workflow chain:

universal-cluster-provision — provisions a new cluster
universal-cluster-destroy — destroys an existing cluster

Cluster workflow chains use universal-start / universal-destroy entry points and execute cluster-specific stages in addition to per-machine tasks.

Cluster vs. Pool¶

Concept	Purpose
Pool	Unallocated resource reservation (which machines are available)
Cluster	Active coordination of allocated machines toward a shared goal

Pools feed machines into clusters: a Resource Broker allocates machines from a pool and the cluster manages their lifecycle together.

Relationship to Resource Brokers¶

Clusters and Resource Brokers work together in the Universal Workflow system. See Universal Broker Architecture for how brokers abstract machine allocation that feeds into clusters.

For the full Universal Workflow system description, see Universal Architecture.

Operational Details¶

For step-by-step instructions on running cluster workflows, see Universal Workflow Cluster Operations.