Fleet, HA and DRS

Add hypervisors by installing machina-agent on each one. The controller then treats them as one pool.

./scripts/deploy-remote.sh USER@HOST --platform --platform-bind 0.0.0.0

What the controller does

• Inventory and desired state. The controller records what should be running where and reconciles drift.
• HA failover. When a host stops responding, its protected VMs are restarted on healthy hosts, with host fencing to avoid split-brain.
• DRS. Distributed resource scheduling rebalances load across hosts by live-migrating VMs.
• Live migration. Move running VMs between hosts from the UI or API.
• Placement. New VMs land on a host chosen by capacity and policy.
• Fleet-wide eBPF. Policies, shield, isolation and telemetry fan out to every host's machina-bpfd through its agent.

How failover, fencing, DRS and multi-controller leader election work in detail: Controller HA.

Daemon-only fleets

Without the controller, machina-daemon can still aggregate peer daemons for a merged inventory:

[fleet]
enabled = true
primary_peer = "hv-east"
standby_peer = "hv-west"

[[fleet.peers]]
name = "hv-east"
url = "https://hv-east.example.com:5092"
api_token = "mach_…"

Endpoint	Purpose
GET /api/v1/fleet/status	Peer health and versions
GET /api/v1/fleet/vms	Merged VM list (local and peers)
POST /api/v1/fleet/peers/{name}/proxy	Forward an API call to a peer

This mode is for inventory and active/standby operation; automatic failover comes from the controller.