List a node's temperature sensors (operator-only).
curl --request GET \ --url https://stratum.example.com/v1/compute/nodes/example/thermals \ --header 'Authorization: Bearer <token>'. Returns the host’s temperature sensors — CPU package, per-core, and any chassis/ambient sensor — each with the critical threshold the hardware itself reports and the headroom remaining to it.
The companion to GET /v1/compute/nodes/{nodeId}/disks, whose health
block covers the drives: together they answer “how is this box” without
the operator having to guess whether a hot disk is a bad drive or a hot
chassis.
Read-only, collected on the same refresh pass as the disk/NIC/DIMM
inventory by reading /sys/class/hwmon. No sensor exists for DIMM
temperature or fan RPM on some hardware, so neither is reported.
Cedar action compute:DescribeNode. Platform-tenant.
Authorizations
Section titled “Authorizations”Parameters
Section titled “Parameters”Path Parameters
Section titled “Path Parameters”Responses
Section titled “Responses”The node’s temperature sensors, in hardware order.
object
One host temperature sensor, read from /sys/class/hwmon.
The sibling of NodeDiskHealth: that one says whether the drives are dying,
this one says whether the box they sit in is being cooled. A drive running
hot because the whole chassis is hot is a different fix from one drive in a
bad slot, and answering that needs both.
Judge a reading by headroomCelsius, not by temperatureCelsius. The
absolute number does not travel between machines: 76 C is unremarkable on a
part whose silicon reports a 100 C critical point and alarming on one that
throttles at 85. The distance to the hardware’s own limit means the same
thing everywhere. This is the same lesson landed on for disks after
finding three different warning thresholds across four drive models.
Note for some hardware: DIMMs expose no thermal sensors (no jc42 / SPD hub), and there are no fan tachometers, so neither RAM temperature nor fan RPM appears here. That is the hardware, not a gap in the collector.
object
Hwmon driver name, e.g. coretemp or acpitz. Suffixed with -N when a board exposes the same chip more than once (dual socket).
Stable per-chip key, slugified from the hardware label (“Package id 0” becomes package_id_0), so it survives hwmon renumbering across reboots. Falls back to tempN for an unlabelled sensor.
The hardware’s own label, verbatim. Empty when it has none.
package— whole-CPU-package temperature. The signal worth alarming on: it is normally the hottest core, and it answers “is this chassis cooling adequately”.core— one physical core. A single hot core while the package is fine points at load placement, not cooling.ambient— chassis/board sensor. Reported because it is free; it carries no thresholds and nothing is built on it.other— anything else the board exposes.
Current reading.
The sensor’s own high/max threshold (tempN_max). Absent when the hardware reports none.
The sensor’s own critical threshold (tempN_crit) — the hardware’s statement about itself, never a Stratum constant. Absent when the hardware reports none.
criticalCelsius - temperatureCelsius: degrees left before the hardware’s own limit, negative once past it. The number to alarm and chart on. Absent when the sensor reports no critical threshold.
Derived from headroomCelsius, and a SNAPSHOT — a package temperature
spikes towards its limit under a burst and means nothing. Sustained heat
is what indicates failing cooling, which is why the node-cpu-hot alarm
requires the condition to hold for ten consecutive minutes rather than
firing on this field directly.
critical— at or past the point the hardware itself calls critical.hot— within 15 C of that point. Throttling territory.healthy— comfortably below it.unknown— the reading exists but the hardware reports no threshold to judge it against (acpitzhere). Never render as healthy: it is an absence of information, not a verdict.
Example
{ "items": [ { "kind": "package", "condition": "unknown" } ]}Resource not found.
object
The findings of a rejected IAM policy document (custom and resource policies).
One finding of a rejected IAM policy document. A rejection lists every finding, so the author fixes the document in one round trip.
object
The finding, e.g. INVALID_DOCUMENT, UNKNOWN_ACTION, ACTION_NOT_APPLICABLE.
Index of the statement at fault; -1 for the document as a whole.
The offending action, where one is at fault (ACTION_NOT_APPLICABLE).
Field errors, for codes that carry them: each problem at its field
(see FieldError). Documented on the responses that set it, e.g.
approveNodeJoin’s INVALID_INPUT and JOIN_APPROVAL_REFUSED.
Absent otherwise.
One field’s problem, the stage-2 installer API’s shape.
field is the field’s path, message what is wrong with it.
object
The field’s path, e.g. links[0].interface, disks[1].confirm,
mgmt.address or role; empty for the whole request.
Structured facts about the refusal, for codes that carry them. The
shape is fixed per code and documented on the response that answers
it, e.g. IMAGE_POOL_MISMATCH (see ImagePoolMismatchDetails) and
IMAGE_IN_USE. Absent for codes that carry none.
object
Example
{ "issues": [ { "code": "ACTION_NOT_APPLICABLE", "action": "compute:CreateVpc" } ]}