> ## Documentation Index > Fetch the complete documentation index at: https://docs.costgraph.ai/llms.txt > Use this file to discover all available pages before exploring further. # Get accelerator metrics > Returns time-series metrics for one accelerator held by the tenant selected by the X-CostGraph-Tenant-ID header, read from the metrics store rather than from the stored reports the detail endpoint returns. Pass mig_instance_id to read one partition of a carved-up card instead of the whole card. gpu_availability states each dimension separately: supported where the card measured it, capacity_unknown where the card measured it but reports no capacity to express it as a percentage, unsupported where the card cannot report it at all, and no_data where nothing was measured; none of the four is ever returned as zero. gpu_memory_used_mib and gpu_power_watts carry the raw measurement behind the memory and power ratios, in mebibytes and watts, whenever the card reported it, so a capacity_unknown dimension still arrives with its numbers rather than as an empty series. gpu_series carries every other signal the card reports as its own time series, keyed by the same names the stored reports use: gpu_temperature, gpu_memory_temperature, gpu_sm_clock, gpu_memory_clock, gpu_encoder_utilization, gpu_decoder_utilization, gpu_memory_copy_utilization, gpu_sm_active, gpu_sm_occupancy, gpu_pcie_tx_bytes, gpu_pcie_rx_bytes, gpu_energy_joules, gpu_xid_errors, gpu_remapped_rows_correctable, gpu_remapped_rows_uncorrectable and gpu_row_remap_failure. A signal the card does not report is left out of gpu_series and named unsupported in gpu_availability, which now states all of them; a card without a memory temperature sensor reports a literal 0, and that is dropped rather than charted as freezing. gpu_units gives the unit of every series, including gpu_memory_used_mib and gpu_power_watts, so a caller never has to infer one: percent, celsius, megahertz, bytes, joules, count, mebibytes or watts. The fault counters and energy are reported as what accrued in the five minutes before each point rather than as the driver's running total, which only ever climbs. gpu_identity_source names the label the series were matched on, since not every card carries a costgraph instance id. A uuid the tenant has never reported gets a 404, as does a mig_instance_id the card has never reported. ## OpenAPI ````yaml /api-reference/costgraph/openapi.json get /api/v1/tenant/gpus/{gpu_uuid}/metrics openapi: 3.0.0 info: description: Read and manage your CostGraph organization, spend, alerts, and settings. title: CostGraph API contact: {} version: '1.0' servers: - url: https://api.costgraph.ai security: [] tags: - name: ai x-group: AI - name: ai-serving x-group: AI serving - name: anomalies x-group: Anomalies - name: auth x-group: Auth - name: billing x-group: Billing - name: billing-export x-group: Billing export - name: budgets x-group: Budgets - name: ci x-group: CI - name: compute-recommendations x-group: Compute recommendations - name: config x-group: Config - name: cost x-group: Cost - name: gpus x-group: GPUs - name: graphai x-group: Graph AI - name: infracost x-group: Infracost - name: integrations x-group: Integrations - name: invitations x-group: Invitations - name: kubernetes-clusters x-group: Kubernetes clusters - name: marketplace x-group: Marketplace - name: network-requests x-group: Network requests - name: notifications x-group: Notifications - name: oauth x-group: OAuth - name: oauth-clients x-group: OAuth clients - name: opencost x-group: OpenCost - name: organization x-group: Audit log - name: organizations x-group: Organizations - name: placement-alternatives x-group: Placement alternatives - name: reports x-group: Reports - name: service-map x-group: Service map - name: settings x-group: Settings - name: sso x-group: Single sign-on - name: tenants x-group: Tenants - name: user x-group: Users - name: virtual-machines x-group: Virtual machines - name: virtual-tags x-group: Virtual tags - name: workflows x-group: Workflows paths: /api/v1/tenant/gpus/{gpu_uuid}/metrics: get: tags: - gpus summary: Get accelerator metrics description: >- Returns time-series metrics for one accelerator held by the tenant selected by the X-CostGraph-Tenant-ID header, read from the metrics store rather than from the stored reports the detail endpoint returns. Pass mig_instance_id to read one partition of a carved-up card instead of the whole card. gpu_availability states each dimension separately: supported where the card measured it, capacity_unknown where the card measured it but reports no capacity to express it as a percentage, unsupported where the card cannot report it at all, and no_data where nothing was measured; none of the four is ever returned as zero. gpu_memory_used_mib and gpu_power_watts carry the raw measurement behind the memory and power ratios, in mebibytes and watts, whenever the card reported it, so a capacity_unknown dimension still arrives with its numbers rather than as an empty series. gpu_series carries every other signal the card reports as its own time series, keyed by the same names the stored reports use: gpu_temperature, gpu_memory_temperature, gpu_sm_clock, gpu_memory_clock, gpu_encoder_utilization, gpu_decoder_utilization, gpu_memory_copy_utilization, gpu_sm_active, gpu_sm_occupancy, gpu_pcie_tx_bytes, gpu_pcie_rx_bytes, gpu_energy_joules, gpu_xid_errors, gpu_remapped_rows_correctable, gpu_remapped_rows_uncorrectable and gpu_row_remap_failure. A signal the card does not report is left out of gpu_series and named unsupported in gpu_availability, which now states all of them; a card without a memory temperature sensor reports a literal 0, and that is dropped rather than charted as freezing. gpu_units gives the unit of every series, including gpu_memory_used_mib and gpu_power_watts, so a caller never has to infer one: percent, celsius, megahertz, bytes, joules, count, mebibytes or watts. The fault counters and energy are reported as what accrued in the five minutes before each point rather than as the driver's running total, which only ever climbs. gpu_identity_source names the label the series were matched on, since not every card carries a costgraph instance id. A uuid the tenant has never reported gets a 404, as does a mig_instance_id the card has never reported. parameters: - description: Tenant ID name: X-CostGraph-Tenant-ID in: header required: true schema: type: string - description: Accelerator UUID as the device reports it name: gpu_uuid in: path required: true schema: type: string - description: Time window (e.g. 1h, 6h, 24h, 7d). Default 1h name: range in: query schema: type: string - description: Resolution step (e.g. 30s, 1m, 5m) name: step in: query schema: type: string - description: 'Metric groups to fetch (comma-separated): all,gpu. Default gpu' name: groups in: query schema: type: string - description: Restrict to one MIG partition of the card name: mig_instance_id in: query schema: type: string responses: '200': description: OK content: application/json: schema: allOf: - $ref: '#/components/schemas/responses.SuccessResponse' - type: object properties: data: $ref: >- #/components/schemas/costgraph_agent.VirtualMachineMetrics '400': description: Bad Request content: application/json: schema: $ref: '#/components/schemas/responses.ErrorResponse' '401': description: Unauthorized content: application/json: schema: $ref: '#/components/schemas/responses.ErrorResponse' '403': description: Forbidden content: application/json: schema: $ref: '#/components/schemas/responses.ErrorResponse' '404': description: Not Found content: application/json: schema: $ref: '#/components/schemas/responses.ErrorResponse' '500': description: Internal Server Error content: application/json: schema: $ref: '#/components/schemas/responses.ErrorResponse' security: - BearerAuth: [] components: schemas: responses.SuccessResponse: type: object required: - message - status properties: data: {} message: type: string example: some message status: type: string example: success costgraph_agent.VirtualMachineMetrics: type: object required: - end - start - step_seconds properties: cpu_irq: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' cpu_nice: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' cpu_psi_some: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' cpu_softirq: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' cpu_steal: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' cpu_system: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' cpu_user: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' cpu_utilization: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' cpu_wait: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' disk_busy_percent: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' disk_read_iops: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' disk_read_latency: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' disk_read_throughput: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' disk_utilization: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' disk_write_iops: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' disk_write_latency: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' disk_write_throughput: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' end: type: string gpu_availability: type: object additionalProperties: type: string gpu_identity_source: type: string gpu_memory_used_mib: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' gpu_memory_utilization: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' gpu_power_utilization: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' gpu_power_watts: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' gpu_series: type: object additionalProperties: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' gpu_units: type: object additionalProperties: type: string gpu_utilization: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' io_psi_some: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' load_average_15m: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' load_average_1m: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' load_average_5m: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' memory_available: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' memory_buffers: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' memory_buffers_percent: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' memory_cached: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' memory_cached_percent: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' memory_free: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' memory_psi_some: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' memory_used: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' memory_utilization: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' network_receive: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' network_transmit: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' start: type: string step_seconds: type: integer swap_utilization: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' tcp_retransmits: type: array items: $ref: '#/components/schemas/metrics.MetricPoint' responses.ErrorResponse: type: object required: - message - status properties: message: type: string example: some message status: type: string example: error metrics.MetricPoint: type: object required: - t - v properties: t: type: integer v: type: number securitySchemes: BearerAuth: description: Enter "Bearer {token}" type: apiKey name: Authorization in: header ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.