quickstart is the Ansible playbook that deploys an Eru cluster on Ubuntu hosts. It installs the metadata store, the scheduler in front of it, and the per-node daemon that reports what the nodes are doing, then registers every node with the scheduler.
inventory.yml
│
┌──────────┼───────────────┬─────────────────┬──────────────────┐
│ │ │ │ │
[etcd] [core] [node_containerd] [node_process] [node_cocoon]
│ │ │ │ │
etcd eru-core containerd systemd cocoon
:2379 │ │ │ │ │
▲ │ └─ ssh ──────►│◄────── ssh ─────┴──── ssh ─────────┘
└─────────┘ eru-agent eru-agent eru-agent
metadata, locks └───────── grpc :5001 ──────────────┘
Core is the only component that talks to etcd. It reaches every node over SSH with its own key
pair — nodes run no daemon that core connects to — and eru-agent reports node and workload status
back over gRPC. The engine behind a node is decided by the endpoint it was registered under:
containerd:// for containers, process:// for bundles that run as transient systemd units, and
cocoon:// for VM workloads on a host that already runs cocoon.
Each inventory group is a role. A host may belong to etcd, core and one node group at once — the
single-node bootstrap puts one host in all three.
Source and issue tracker: github.com/projecteru2/quickstart. Part of the Eru cluster stack, alongside core, agent, cli and resource-extend.