Infrastructure
Infrastructure is the physical layer backing the cluster: racks, the chassis mounted in them, and the compute nodes (and their processing units) that live in chassis blades. The Infrastructure page renders the whole physical layout — a list of racks, a toggle to view the compute-node layer, the physical-disk layer, or both, and a click-to-inspect panel for any node or disk slot.
For specific tasks, see Configure Rack and Configure Chassis Type under this section.
Draft — needs screenshots. This section hasn't been through the docs-capture pipeline yet, so there are no screenshots. Steps are grounded in the current product code (routes and page components), not a live capture — verify against the running product once this section is captured.
Browsing the physical layout
The top of the page shows cluster-wide totals — number of racks, number of chassis, number of compute nodes, and used/total storage, memory, and cores across the whole cluster. The same page also exposes a cluster-wide High Availability toggle (enabling it requires at least 3 compute nodes); that's a clustering setting rather than a physical-infrastructure control, so it isn't covered further here.
Below that, a layer toggle — Show all Layers, Show Compute Layer, or Show Storage Layer — controls whether the rack diagrams render compute node slots, physical disk slots, or both. A search box filters the rack view down to compute nodes whose ID matches the search term.
Racks are listed as an accordion; each one renders its mounted chassis at their assigned rack-unit position. If any chassis exist that haven't been mounted to a rack yet, they're grouped under a synthetic "Unassigned Chassis" entry with a link into the rack configurator to mount them.
Clicking a compute node slot opens an info panel showing:
- The node's rack ID, chassis ID, and blade position (or blade ID).
- A Details link to the node's full detail page (see below).
- The list of processing units hosted on that node, each showing its type (for example, a failover-primary unit is labeled accordingly), a status (Active, Error, Reset, Initializing Reboot, Removed, or another state shown as a general warning), allocated/total cores, and allocated/total memory. Per-unit actions include Reboot and Remove Node (or Activate Node for a previously removed unit).
Clicking a physical disk slot opens an equivalent panel for that disk.
Compute node detail
Following a node's Details link (or navigating to
/infrastructure/nodes/:id directly) opens the node's detail page: a header
with the node's label, vendor, rack ID, and processing-unit count, and three
tabs.
CPU Usage (the default tab) shows an Allocated Cores widget — the overall percentage of the node's cores currently assigned across resource groups — next to a CPU Groups panel. In edit mode, each core (grouped by socket) can be assigned to, or removed from, a resource group; changes are staged and applied with Save, or discarded with Cancel. Below that, per-core usage bar charts are broken out by socket, plus a chart of CPU usage by instance. Resource groups themselves are a separate concern, covered under Resource Groups.
Storage shows the node's used/total storage and physical disk count, with filters for disk state, datastore, and resource group, rendering the matching disks as cards.
Instances lists the VM instances running on the node's processing units, with node-level progress bars for in-flight operations. While this tab is active, the page header also offers Startup all / Shutdown all actions for every instance on the node — each is only available when all of the node's instances share a compatible state (for example, "Startup all" requires them all to already be stopped).
Neither this page nor the rack browser above surfaces dedicated "isolation" or "PCI passthrough" controls, despite those terms showing up in general platform descriptions of node management. CPU allocation here is expressed as core-to-resource-group assignment (above); passthrough of specific PCI devices, if configurable, would live at the level of an instance or a resource group, not here.