Skip to content
StratumStratum Docs

Hyperconverged cluster

A cluster is built one machine at a time. The first node, the founder, installs as a cluster of one. Every other machine boots the same ISO, registers with a join code and joins once you approve it. Each node runs the control plane, the workloads and, when you give it disks, a Ceph OSD.

Install a clusterPlan the nodes and networks, found the cluster, join the others, set up storage.
Manage a clusterHealth, adding and removing nodes, shutdown, failures, backup, upgrades and recovery.

For one machine, see Install a single node.

What a cluster adds

Section titled “What a cluster adds”
  • No single point of failure. Postgres, OpenBao and the OVN databases run on every control-plane node and elect leaders. Each service has an address that follows its leader.
  • Replicated storage. Ceph keeps three copies of each volume. An instance on Ceph restarts on another node when its node dies.
  • More services. Object storage, managed databases, Kubernetes and Auto Scaling need a cluster.

The steps

Section titled “The steps”
  1. Before you start: nodes, disks, networks.
  2. Install the first node.
  3. Create a join code, install the other nodes and approve them.
  4. Check readiness, then set up storage.
  5. Build the operator images for managed databases and Kubernetes.
  6. Verify.

Installing, signing in, requirements, firmware, the USB stick and upgrade paths are shared with a single node: requirements, firmware, USB stick, download, first sign-in and upgrades.

API calls on these pages

Section titled “API calls on these pages”

Some tasks have no stratumctl command yet, most of Ceph among them. Those pages call the API with curl, and name the console page that does the same. $API is the cluster’s API address (http://192.0.2.240:8443, the API VIP) and $TOKEN is the token of your profile in ~/.stratum/config.yaml, saved by stratumctl login.