Nodes and hardware

Every node reports its own hardware with each heartbeat: CPU, memory, disks and their health, NICs, PCI devices and temperatures. Nothing here is probed from another node.

The node list

Console: Admin › Metal › Overview. One row per node with its role, state, committed vCPU and memory against what it can offer, and when it last reported.

stratumctl nodes list
stratumctl nodes list --state cordoned
stratumctl nodes get n-abc123

The node list with commitment and health.

On a cluster, machines waiting to join show under Joining nodes, and Add node makes a join code. See Add a node, and Remove or revoke a node for taking one out.

Field	Meaning
Role	`hybrid` runs the control plane and instances (every single node is hybrid); `hypervisor` runs instances only.
State	`ready`, `cordoned` (no new instances), `draining`, `maintenance`, or `unknown` (no recent heartbeat).
Reserved / Allocatable	What the node keeps for the platform, and what that leaves for instances. See Host reservation.
Last seen	The last heartbeat. A node silent for long is `unknown`, and readiness reports `NodeNotReporting`.

A node’s detail

Each node has its own entry under Metal in the console, with these tabs:

Tab	Shows
Overview	Name, ID, hostname, role, region, availability zone, architecture, last seen; host metrics over 15 minutes to 24 hours.
Labels	Topology (region, zone, rack), detected capabilities, operator labels and taints that steer the scheduler.
CPU	Model, sockets, cores, threads; virtualization flags (VT-x or AMD-V, nested virtualization), kernel and KVM; temperatures and their headroom.
Memory	Total, used and committed memory; the memory modules the firmware reports.
Disks	Every block device with its use (OS, pool, free) and drive health; filesystems; the raw LVM view.
Network	NICs with link, speed, MAC and addresses, and the node’s networks.
Devices	PCI devices, GPUs and accelerators first, with SR-IOV capacity.
Compute	The instances placed on the node.

Firmware facts (UEFI, Secure Boot) show on the Overview tab.

Drive health

Each drive’s SMART data rolls up into one condition, worst first:

Condition	Means	Alarm
`failing`	Reliability is degraded: spare exhausted, media errors, reallocated or pending sectors, or the drive reports itself degraded. Replace it.	`disk-failing`
`wearing`	Flash endurance used past the drive’s bound. Weeks to months of notice.	`disk-wearing`
`hot`	Above the drive’s own warning temperature. A cooling problem.	`disk-hot`
`healthy`	SMART read, nothing wrong.
`unknown`	SMART could not be read. This is missing information, not a healthy drive.

The reason names the drive’s own threshold, for example “72C is above this drive’s 70C warning threshold”. See Disks to replace one.

stratumctl nodes disks n-abc123

Other hardware alarms

Alarm	Fires when
`node-cpu-hot`, `node-cpu-critical`	CPU temperature near or past its limit.
`node-nic-flapping`	A NIC’s link keeps going up and down.
`node-disk-low`	A node filesystem is running out of space.
`md-raid-degraded`	An MD array lost a member.
`host-resource-pressure`	The host is short of memory, CPU or I/O.

They go to the operator’s notification channel; see Alarms and notifications.