Spec Metrics

Feature Overview

Item	Content
Applicable Role	Operator
Navigation Path	AI Infra(On-Prem) > Resource Pools > Spec Metrics
Page Route	/powerone/resourcepool/flavor/type
Managed Object	Configuration, status, and relationships on Spec Metrics

Beginner Explanation

Specification metrics are like units of measure in a resource specification table. Whether CPU, memory, GPU, VRAM, and other fields can be correctly identified, displayed, and counted depends on metric definitions. When metric definitions are inconsistent, the specification name users see can fail to match the actual scheduled resources.

Terms

Term	Description
Metric Name	Metric name displayed on the page and referenced by resource specifications.
Metric Type	CPU, memory, AI accelerator metric, or another metric type provided by the page.
Resource Key	Resource identifier used for scheduling or metering.

Recommended Operation Order

Confirm prerequisites for Metric name, metric type, resource key, unit, k8s-key, selector-key, monitoring metric, and enabled status, follow Main Operations, run Result Validation, and continue to the next page.

First-Time User Notes

Confirm that the task involves Configuration, status, and relationships on Spec Metrics, and then follow the recommended order. If fields or state differ from expectations, check prerequisites before continuing downstream.

Prerequisites

1. The current account has operator permissions and can access AI Infra > On-Prem > Resource Pools > Specification Metrics.
2. Target cluster resource reporting definitions have been confirmed, including k8s-key, selector-key, unit, and monitoring metric mapping.
3. If creating an AI accelerator metric, accelerator model, node labels, and device plugin reporting information have been confirmed.
4. The impact on resource specifications, templates, job scheduling, or metering rules has been evaluated.

Page Description

Use this page to manage metric names, metric types, resource keys, units, Kubernetes keys, selector keys, monitoring metrics, and enabled state.

The page displays configured metrics as cards and supports filtering by metric name, AI accelerator metrics, and other metrics.

The following figure shows the specification metric list, where monitoring metrics, k8s-key, selector-key, and units can be viewed.

Main Operations

View Specification Metrics

1. Go to Resource Pools > Flavor Metrics.
2. Filter by name, type, unit, status, or update time.
3. Open details and check data type, unit, range, default, and referenced flavors.
4. If no record is returned, reset filters. For inconsistent units or ranges, check references first.

Add Specification Metric

Applicable Scenarios

Add a specification metric when a new hardware resource type needs to be added, CPU or memory metrics need maintenance, or an AI accelerator needs to be connected to resource specifications and job scheduling.

Steps

1. Go to AI Infrastructure > On-Prem > Resource Pools > Specification Metrics.
2. Click "Add" or the actual add entry on the page.
3. Select a metric type, such as CPU, memory, AI accelerator metric, or another metric type provided by the page.
4. Select AI card category, AI cards metrics, or other metrics, and fill in metric name, unit, k8s-key, and selector-key according to the page fields.
5. For an AI accelerator metric, verify that selector-key is consistent with labels actually reported by cluster nodes.
6. Before clicking the final "Save", "Submit", or "OK", verify the metric scope, unit, k8s-key, and selector-key again.

The following figure shows the Add Specification Metric drawer. AI accelerator metrics require k8s-key and selector-key.

Import or Export Specification Metrics

Applicable Scenarios

Use the "Import/Export" menu to batch-maintain specification metrics, or to export the current metric definitions for audit, reconciliation, and controlled migration.

Steps

1. Go to AI Infrastructure > On-Prem > Resource Pools > Specification Metrics.
2. Click "Import/Export" and choose "Import" or "Export" according to the business purpose.
3. For import, upload the file as required by the page and verify metric type, name, unit, k8s-key, and selector-key.
4. For export, confirm the filter scope and generate and download the metric list as prompted by the page.
5. Before importing, verify that an incorrect metric used by resource specifications will not be overwritten. Save export files in a controlled directory.

Result Validation

• After import, the specification metrics list shows the added or updated definitions.
• The metric scope in the export file matches the current filter conditions.
• The Resource Specifications page can correctly reference the imported metrics.

The import dialog uses "Read File" to upload the ZIP configuration package exported by the backend and run a pre-check. Before submitting, verify that the metric type, name, unit, k8s-key, and selector-key match the target environment.

The export dialog selects the scope under "Select Module" by groups such as resource pool, templates, and VRAM calculation. Before exporting, confirm the current filter conditions to avoid bringing out unnecessary metric definitions.

Notes

• Unit, k8s-key, and selector-key affect resource reporting, specification matching, and scheduling. Do not judge equivalence by display name alone.
• Verify identifiers and reference relationships before importing to avoid overwriting an in-use metric definition.

Parameter Quick Reference

Field Name	Required	Field Type	Example	Description
AI card Category	Conditionally required	Dropdown / enum	CPU	Category selected when adding an AI accelerator metric. Keep it consistent with hardware categories in Accelerators.
AI Cards Metrics	Conditionally required	Dropdown / enum	Example value	Metric type selected for AI accelerator spec metrics. Used to distinguish accelerator metrics.
Other Metrics	Conditionally required	Port / number	CPU	CPU, memory, or another non-accelerator metric supported by the page. Keep it consistent with resource specification display and scheduling scope.
Metric Name	Yes	Text	GPU VRAM	Display name of the specification metric. Use a name that expresses resource type and unit.
Unit	Yes	Text	GiB	Display or metering unit. Keep it consistent with capacity statistics and Resource Specs.
k8s-key	Conditionally required	Credential / sensitive text	nvidia.com/gpu	Kubernetes scheduling resource key. Must match the resource key actually reported by Kubernetes nodes.
selector-key	Conditionally required	Credential / sensitive text	accelerator	Accelerator model, node label, or device selector key. Must match the accelerator model or node label.
Actions	System-generated	Action entry	Edit	Add, edit, import/export, delete, and similar entries. Confirm submits real configuration. Confirm the scope and impact before executing the final action.

Pitfalls

• Adding specification metrics affects resource specifications, job scheduling, monitoring display, and metering definitions.
• k8s-key must match the resource key actually reported by Kubernetes nodes, or jobs may fail to request resources.
• selector-key must match accelerator models or node labels, or AI accelerator metrics may fail to match devices correctly.
• Incorrect metric units may cause specification display, capacity statistics, or template recommendation deviations.
• Before disabling or deleting metrics referenced by resource specifications, confirm the impact on specifications, templates, and running jobs.
• Save, Submit, and OK are high-risk final actions. Confirm the scope and impact before executing the final action.

Configuration Rules and Impact

• Metric before specification: Resource specifications must reference existing and available specification metrics.
• k8s-key consistency: Use the key actually reported by Kubernetes nodes, not the page display name.
• selector-key consistency: The selector-key of AI accelerator metrics should be consistent with accelerator models or node labels.
• Unit consistency: Capacity metrics such as memory, VRAM, and disk should use unified GiB, GB, or platform-defined units.
• Disable impact: Disabling a metric may affect resource specifications, template recommendations, job creation, and metering statistics.
• Monitoring impact: Incorrect monitoring metric mapping may cause abnormal resource monitoring display or capacity statistics deviations.

Result Validation

Check Item	Success Signal	If Abnormal
Page entry	Spec Metrics opens with the target operation entry	Check Operator permission and whether the menu is available
Object record	Configuration, status, and relationships on Spec Metrics is visible in the list or details	Reset filters and verify name, ownership, and creation result
State result	State after creation or change matches the page message	Check operation feedback, dependency state, and latest update time
Downstream use	A downstream page can select or associate the target	Return to prerequisites and check enabled state, ownership, and visibility

FAQ

An Imported Metric Cannot Be Referenced by Resource Specifications

Symptom:

The metric import completes, but the metric cannot be found when creating or editing a resource specification.

Possible Causes:

• The metric is disabled or still being validated.
• Metric type, unit, or identifier does not match.
• The current account cannot see the resource specification scope.

Solution:

1. Check the metric state, type, unit, and update time in Specification Metrics.
2. Recheck the import file against the metric type and resource key used by Resource Specifications.
3. Confirm visibility scope and refresh the resource specification form.

Target Is Missing from Spec Metrics

Symptom:

The page opens, but the expected Configuration, status, and relationships on Spec Metrics is missing.

Possible Causes:

• Filters remain active.
• the object belongs to another scope.
• a prerequisite is incomplete.

Solution:

1. Reset filters
2. verify region or tenant ownership
3. confirm prerequisite state.

The Operation Entry on Spec Metrics Is Unavailable

Symptom:

The create, register, or maintain entry is hidden or disabled.

Possible Causes:

• Role permission is insufficient.
• the page is read-only.
• dependencies are not ready.

Solution:

1. Check Operator permission
2. read the page message
3. complete dependency configuration first.

Spec Metrics Has an Abnormal State After the Operation

Symptom:

A record exists after submission, but its state is unexpected.

Possible Causes:

• Connectivity or validation failed.
• a dependency is abnormal.
• processing is incomplete.

Solution:

1. Check feedback and update time
2. inspect related objects
3. troubleshoot the processing stage.

A Downstream Page Cannot Use Spec Metrics

Symptom:

The current page is normal, but a downstream page cannot select or associate Configuration, status, and relationships on Spec Metrics.

Possible Causes:

• Visibility differs.
• the object is disabled.
• downstream cache is stale.

Solution:

1. Check enabled state and ownership
2. verify role visibility
3. refresh and select again.

Notes

• Adding specification metrics affects resource specifications, job scheduling, monitoring display, and metering definitions.
• k8s-key, selector-key, unit, and monitoring metric mapping must follow real cluster and platform definitions.
• Do not directly delete metrics referenced by specifications. Migrate specifications and verify job scheduling first.
• Save, Submit, and OK are high-risk final actions.
• Do not write real internal resource key mappings, node labels, cluster IDs, resource pool IDs, internal addresses, accounts, keys, tokens, or internal test parameters.

Next Steps

1. Go to Resource Pools > Resource Specifications to create or adjust specifications.
2. Go to Resource Pools > Accelerator Management to confirm accelerator model associations.
3. Verify in a test job that the metric resource can be requested normally.
4. Return to the Specification Metrics list and confirm that enabled status, unit, and filter results are as expected.