Prometheus Metrics Interface

Function Description

Provides a Metrics interface for Prometheus to call and integrate.

URL

GET http://ip:port/metrics

For security reasons, NPU Exporter enables the container-level port (default 8082) by default. The request IP is the Kubernetes container IP. When the K8s network plugin is Calico, the network policy is set to allow access from applications with the label app=prometheus.

Request Parameters

None

Response Description

Data is returned in the Prometheus-specific format. The relevant data information is shown below for reference only, and the actual echo shall prevail. For detailed description of the data information, see the following text or Data Information Description.xlsx. The data information that comes with Prometheus does not need attention and is not shown here. Some data information is only supported by certain product forms, and the actually reported data information shall prevail.

...

# HELP machine_npu_nums Amount of npu installed on the machine.
# TYPE machine_npu_nums gauge
machine_npu_nums 8
# HELP machine_card_nums Amount of card installed on the machine.
# TYPE machine_card_nums gauge
machine_card_nums 8
# HELP npu_chip_info_aicore_current_freq the npu ai core current frequency, unit is 'MHz'
# TYPE npu_chip_info_aicore_current_freq gauge
npu_chip_info_aicore_current_freq{container_name="",id="0",model_name="910A-Ascend-V1",namespace="",pcie_bus_info="0000:61:00.0",pod_name="",vdie_id="185011D4-21104518-A0C4ED94-14CC040A-56102003"} 1000 1723621883587
npu_chip_info_aicore_current_freq{container_name="",id="1",model_name="910A-Ascend-V1",namespace="",pcie_bus_info="0000:DB:00.0",pod_name="",vdie_id="185011D4-21E04718-93B2ED94-14CC040A-BF102003"} 1000 1723621883932
npu_chip_info_aicore_current_freq{container_name="",id="2",model_name="910A-Ascend-V1",namespace="",pcie_bus_info="0000:B2:00.0",pod_name="",vdie_id="185011D4-20C02418-59D4ED94-14CC040A-F9102003"} 1000 1723621884277
npu_chip_info_aicore_current_freq{container_name="",id="3",model_name="910A-Ascend-V1",namespace="",pcie_bus_info="0000:3E:00.0",pod_name="",vdie_id="185011D4-21502C18-0464ED94-14CC040A-6E102003"} 1000 1723621884682
npu_chip_info_aicore_current_freq{container_name="",id="4",model_name="910A-Ascend-V1",namespace="",pcie_bus_info="0000:60:00.0",pod_name="",vdie_id="185011D4-21A02418-64946D94-14CC040A-F8102003"} 1000 1723621885026
npu_chip_info_aicore_current_freq{container_name="",id="5",model_name="910A-Ascend-V1",namespace="",pcie_bus_info="0000:DA:00.0",pod_name="",vdie_id="185011DC-21F02B18-C4B66D94-14CC040A-57102003"} 1000 1723621885385
npu_chip_info_aicore_current_freq{container_name="",id="6",model_name="910A-Ascend-V1",namespace="",pcie_bus_info="0000:B1:00.0",pod_name="",vdie_id="185011D4-20602118-14646D94-14CC040A-8A102003"} 1000 1723621885784
npu_chip_info_aicore_current_freq{container_name="",id="7",model_name="910A-Ascend-V1",namespace="",pcie_bus_info="0000:3D:00.0",pod_name="",vdie_id="185011D4-21504C18-10656D94-14CC040A-6B102003"} 1000 1723621886131
# HELP npu_chip_info_bandwidth_rx the npu interface receive speed, unit is 'MB/s'
# TYPE npu_chip_info_bandwidth_rx gauge
npu_chip_info_bandwidth_rx{container_name="",id="0",model_name="910A-Ascend-V1",namespace="",pcie_bus_info="0000:61:00.0",pod_name="",vdie_id="185011D4-21104518-A0C4ED94-14CC040A-56102003"} 0 1723621883587
...

This interface supports querying default metric groups and custom metric groups. For details on customizing metric groups, see Custom Metric Development. The default metric groups include the following sections. The collection and reporting of metric groups are controlled by switches in the configuration file. If a switch is configured as enabled, the corresponding metric group will be collected and reported; if a switch is configured as disabled, the corresponding metric group will not be collected and reported.

Version Data Information
Basic Node Information
NPU Data Information
Utilization Data Information
vNPU Data Information
Network Data Information
DDR Data Information
On-Chip Memory Data Information
HCCS Data Information
PCIe Data Information
RoCE Data Information
SIO Data Information
Optical Module Data Information
UB Data Information

To understand the description of labels in the above data information, see Data Label Information.

If the process runs on the host and the Pod does not use an NPU, the values of pod_name, container_name, and namespace will be empty.
NPU Exporter obtains the corresponding information by calling the underlying HDK interface. For the HDK interfaces called for data information, refer to HDK Interface Call.
If NPU Exporter does not support the product form or fails to call the HDK interface when querying a certain data information item, this data information will not be reported.

Version Data Information

Table 1 Version data information

Category	Data Information Name	Data Information Description	Data Label	Field Type	Unit	Supported Product Form
Version	npu_exporter_version_info	NPU Exporter version information	exporterVersion: Current NPU Exporter version information	String	1: Placeholder character, having no actual meaning.	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products Inference Server (with Atlas 300I Inference Card) Atlas Inference Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack Atlas 350 PCIe Card Atlas 850 Series Hardware Atlas 950 SuperPoD

Basic Node Information

Table 2 Basic node information

Category	Data Information Name	Data Information Description	Data Label	Field Type	Unit	Supported Product Form
nodeBase	node_base_info	Basic node Information	exporterVersion: Current NPU Exporter version information driverVersion: Driver version information	String	1: Placeholder character, having no actual meaning.	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products Inference Server (with Atlas 300I Inference Card) Atlas Inference Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack Atlas 350 PCIe Card Atlas 850 Series Hardware Product Atlas 950 SuperPoD

NPU Data Information

Table 3 NPU data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
NPU	machine_card_nums	Number of Ascend AI Processor modules	-	Unit: pcs	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products Inference Server (with Atlas 300I Inference Card) Atlas Inference Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack
NPU	machine_npu_nums	Number of Ascend AI Processors	-	Unit: count	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products Inference Server (with Atlas 300I Inference Card) Atlas Inference Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack Atlas 350 PCIe Card Atlas 850 Series Hardware Atlas 950 SuperPoD
NPU	npu_chip_info_voltage	Ascend AI Processor voltage	Label 1	Unit: Volt (V)
NPU	First error code: npu_chip_info_error_code Other error codes: npu_chip_info_error_code_X	Ascend AI Processor error code When there is no error code on the Ascend AI Processor, this field will not be reported. Note: Prometheus scenario: If multiple error codes exist on the Ascend AI Processor simultaneously, due to Prometheus format limitations, only the first ten error codes that occur are currently reported. The value range of X is 1 to 9. Telegraf scenario: A maximum of 128 error codes can be reported. For detailed descriptions of error codes, see the Chip Fault Code References section.	Label 1	-
NPU	npu_chip_info_name	Ascend AI Processor name and ID	Label 1	-
NPU	npu_chip_info_health_status	Ascend AI Processor health status	Label 1	The value is 0, 1, or -1. 1: Healthy 0: Unhealthy -1: Unknown (DCMI call failed)
NPU	npu_chip_info_temperature	Ascend AI Processor temperature	Label 1	Unit: Celsius (°C)
NPU	npu_chip_info_process_info	Information about processes occupying the Ascend AI Processor Prometheus scenario: The value is the memory used by the process. Telegraf scenario: Reported only when no process occupies the Ascend AI Processor, with a value of 0.	Label 2	Unit: MB
NPU	npu_chip_info_process_info_num	Number of processes occupying the Ascend AI Processor	Label 1	Unit: Count
NPU	npu_container_info	NPU container information Note: Telegraf does not support reporting this metric.	Label 5	-
NPU	npu_chip_info_power	Ascend AI Processor power consumption Note: Only the Atlas Inference Series products report board power consumption; other products report Ascend AI Processor power consumption.	Label 1	Unit: Watt (W)
NPU	npu_chip_info_aicore_current_freq	Current frequency of the Ascend AI Processor's AICore	Label 1	Unit: MHz
NPU	npu_chip_info_serial_number	Ascend AI Processor serial number	Label 6	1: Placeholder character, having no actual meaning
NPU	npu_chip_info_network_status	Ascend AI Processor's network health status	Label 1	The value is 0, 1, or -1. 1: Healthy, reachable 0: Unhealthy, unreachable -1: Unknown (DCMI call failure)	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products A200I A2 Box Heterogeneous Subrack Atlas 800I A2 Inference Server
NPU	container_npu_total_memory	Total NPU memory size with container information Note: Telegraf does not support reporting this metric.	Label 1	Unit: MB	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products A200I A2 Box Heterogeneous Subrack Atlas 800I A2 Inference Server Atlas 350 PCIe Card Atlas 850 Series Hardware Atlas 950 SuperPoD
NPU	container_npu_used_memory	Used NPU memory with container information Note: Telegraf does not support reporting this metric.	Label 1	Unit: MB
NPU	npu_chip_info_product_type	Ascend AI Processor product form	Label 7	1: Placeholder character, having no actual meaning.	Atlas Inference Series Products

Utilization Data Information

Table 4 Utilization data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
utilization	npu_chip_info_utilization	Ascend AI processor's AICore utilization	Label 1	%	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products Inference Server (with Atlas 300I Inference Card) Atlas Inference Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack Atlas 350 PCIe Card Atlas 850 Series Hardware Atlas 950 SuperPoD
utilization	container_npu_utilization	AICore utilization of NPU with container information Note: Telegraf does not support reporting this metric	Label 1	%	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products Inference Server (with Atlas 300I Inference Card) Atlas Inference Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack Atlas 350 PCIe Card Atlas 850 Series Hardware Atlas 950 SuperPoD
utilization	npu_chip_info_vector_utilization	Ascend AI processor's AIVector utilization	Label 1	%	Atlas A2 Training Series Products Atlas A3 Training Series Products Inference Server (with Atlas 300I Inference Card) Atlas Inference Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack Atlas 350 PCIe Card Atlas 850 Series Hardware Atlas 950 SuperPoD
utilization	npu_chip_info_cube_utilization	Ascend AI processor's AICube utilization	Label 1	%	Atlas A2 Training Series Products Atlas A3 Training Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack Atlas 350 PCIe Card Atlas 850 Series Hardware Atlas 950 SuperPoD
utilization	npu_chip_info_overall_utilization	Overall utilization of the Ascend AI Processor	Label 1	%	Atlas A2 Training Series Products Atlas A3 Training Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack Atlas 350 PCIe Card Atlas 850 Series Hardware Atlas 950 SuperPoD

vNPU Data Information

Table 5 vNPU data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
vNPU	vnpu_pod_aicore_utilization	AICore utilization of the vNPU	Label 3	Unit: %	Atlas Inference Series Products
vNPU	vnpu_pod_total_memory	Total memory owned by the vNPU	Label 3	Unit: KB
vNPU	vnpu_pod_used_memory	Used vNPU memory	Label 3	Unit: KB

Network Data Information

Table 6 Network data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
Network	npu_chip_info_bandwidth_rx	Ascend AI Processor's real-time network port RX rate	Label 1	Unit: MB/s	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack
Network	npu_chip_info_bandwidth_tx	Ascend AI Processor's real-time network port TX rate	Label 1	Unit: MB/s
Network	npu_chip_info_link_status	Ascend AI Processor's network port link status	Label 1	The value is 0, 1, or -1. 1: UP 0: DOWN -1: Unknown (hccn_tool call failed)
Network	npu_chip_link_speed	Ascend AI Processor's default network port speed	Label 1	Unit: MB/s
Network	npu_chip_link_up_num	Ascend AI Processor's network port UP statistics	Label 1	Unit: times
Network	npu_chip_info_link_status_X_Y	Link status of the Ascend AI Processor port. X is the Udie ID and Y is the Port ID.	Label 1	The value is 0, 1, or -1. 1: UP 0: DOWN -1: Unknown (hccn_tool invocation failed)	Atlas 350 PCIe Card (4-Processor mesh) Atlas 850 Series Hardware Atlas 950 SuperPoD
Network	npu_chip_info_bandwidth_rx_X_Y	Real-time RX rate of the Ascend AI Processor port. X is the Udie ID and Y is the Port ID. The -time parameter used is 100.	Label 1	Unit: MB/s
Network	npu_chip_info_bandwidth_tx_X_Y	Real-time TX rate of the Ascend AI processor port. X is the Udie ID and Y is the Port ID. The -time parameter used is 100.	Label 1	Unit: MB/s
Network	npu_chip_info_link_speed_X_Y	Speed of the physical port. X is the Udie ID and Y is the Port ID.	Label 1	Unit: G

DDR Data Information

Table 7 DDR data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
DDR	npu_chip_info_used_memory	Used DDR memory of the Ascend AI Processor	Label 1	Unit: MB	Atlas Training Series Products Inference Server (with Atlas 300I Inference Card) Atlas Inference Series Products
DDR	npu_chip_info_total_memory	Ascend AI Processor's total DDR memory	Label 1	Unit: MB

On-Chip Memory Data Information

Table 8 On-chip memory data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
On-Chip Memory	npu_chip_info_hbm_used_memory	Used on-chip memory of the Ascend AI Processor	Label 1	Unit: MB	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products A200I A2 Box Heterogeneous Subrack Atlas 800I A2 Inference Server Atlas 350 PCIe Card Atlas 850 Series Hardware Atlas 950 SuperPoD
On-Chip Memory	npu_chip_info_hbm_total_memory	Ascend AI Processor's total on-chip memory	Label 1	Unit: MB
On-Chip Memory	npu_chip_info_hbm_ecc_enable_flag	On-chip memory ECC enabling status of the Ascend AI Processor	Label 1	The value is 1 or 0. 0: ECC detection is not enabled. 1: ECC detection is enabled.
On-Chip Memory	npu_chip_info_hbm_ecc_single_bit_error_cnt	Single‑bit error count of the Ascend AI Processor's on‑chip memory	Label 1	-
On-Chip Memory	npu_chip_info_hbm_ecc_double_bit_error_cnt	Double‑bit error count of the Ascend AI Processor's on‑chip memory	Label 1	-
On-Chip Memory	npu_chip_info_hbm_ecc_total_single_bit_error_cnt	Total number of all single‑bit errors over the lifetime of the Ascend AI Processor's on‑chip memory	Label 1	-
On-Chip Memory	npu_chip_info_hbm_ecc_total_double_bit_error_cnt	Total number of all double‑bit errors over the lifetime of the Ascend AI Processor's on‑chip memory	Label 1	-
On-Chip Memory	npu_chip_info_hbm_ecc_single_bit_isolated_pages_cnt	Number of memory pages isolated due to single‑bit errors in the Ascend AI Processor's on‑chip memory	Label 1	-
On-Chip Memory	npu_chip_info_hbm_ecc_double_bit_isolated_pages_cnt	Number of memory pages isolated due to double‑bit errors in the Ascend AI Processor's on‑chip memory	Label 1	-
On-Chip Memory	npu_chip_info_hbm_utilization	On-chip memory utilization	Label 1	Unit: %
On-Chip Memory	npu_chip_info_hbm_temperature	On-chip memory temperature	Label 1	Unit: °C
On-Chip Memory	npu_chip_info_hbm_bandwidth_utilization	Bandwidth utilization of the on-chip memory	Label 1	Unit: %

HCCS Data Information

Table 9 HCCS data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
HCCS	npu_chip_info_hccs_statistic_info_tx_cnt_X X range: 1~7 (Atlas A2 Training Series Products or Atlas 900 A3 SuperPoD), 2~7 (Atlas 9000 A3 SuperPoD)	Number of packets sent on the X-th HDLC link, in flits. -1 is reported if collection fails.	Label 1	-	Atlas A2 Training Series Products Atlas A3 Training Series Products
HCCS	npu_chip_info_hccs_statistic_info_rx_cnt_X X range: 1~7 (Atlas A2 Training Series Products or Atlas 900 A3 SuperPoD), 2~7 (Atlas 9000 A3 SuperPoD)	Number of packets received on the X-th HDLC link, in flits. -1 is reported if collection fails.	Label 1	-
HCCS	npu_chip_info_hccs_statistic_info_crc_err_cnt_X X range: 1~7 (Atlas A2 Training Series Products, Atlas 900 A3 SuperPoD), 2~7 (Atlas 9000 A3 SuperPoD)	CRC error in received packets on the X‑th HDLC link, in flits. -1 is reported upon collection failure.	Label 1	-
HCCS	npu_chip_info_hccs_bandwidth_info_profiling_time	HCCS link bandwidth sampling duration. Value range: 1 to 1000	Label 1	Unit: ms
HCCS	npu_chip_info_hccs_bandwidth_info_total_tx	Total TX bandwidth of the HCCS link. -1 is reported if colelction fails.	Label 1	Unit: GB/s
HCCS	npu_chip_info_hccs_bandwidth_info_total_rx	Total RX bandwidth of the HCCS link. -1 is reported if collection fails.	Label 1	Unit: GB/s
HCCS	npu_chip_info_hccs_bandwidth_info_tx_X X range: 1~7 (Atlas A2 Training Series Products, Atlas 900 A3 SuperPoD), 2~7 (Atlas 9000 A3 SuperPoD)	TX data bandwidth for a single HCCS link. -1 is reported if collection fails.	Label 1	Unit: GB/s
HCCS	npu_chip_info_hccs_bandwidth_info_rx_X X range: 1~7 (Atlas A2 Training Series Products, Atlas 900 A3 SuperPoD), 2~7 (Atlas 9000 A3 SuperPoD)	RX data bandwidth of a single HCCS link. -1 is reported if collection fails.	Label 1	Unit: GB/s

PCIe Data Information

Table 10 PCIe data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
PCIe	npu_chip_info_pcie_rx_p_bw	PCIe bandwidth for remote writes received by the Ascend AI Processor	Label 4	Unit: MB/ms	Atlas A2 Training Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack
PCIe	npu_chip_info_pcie_rx_np_bw	PCIe bandwidth for remote reads received by the Ascend AI Processor	Label 4	Unit: MB/ms
PCIe	npu_chip_info_pcie_rx_cpl_bw	PCIe bandwidth for receiving CPL responses to remote reads by the Ascend AI Processor	Label 4	Unit: MB/ms
PCIe	npu_chip_info_pcie_tx_p_bw	PCIe bandwidth for remote writes from the Ascend AI Processor	Label 4	Unit: MB/ms
PCIe	npu_chip_info_pcie_tx_np_bw	PCIe bandwidth for remote reads from the Ascend AI Processor	Label 4	Unit: MB/ms
PCIe	npu_chip_info_pcie_tx_cpl_bw	PCIe bandwidth for the Ascend AI Processor to send CPL replies to remote read operations	Label 4	Unit: MB/ms

RoCE Data Information

Table 11 RoCE data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
RoCE	npu_chip_mac_rx_pause_num	Total number of Pause frames received by MAC	Label 1	-	Atlas Training Series Products Atlas A2 Training Series Products Atlas A3 Training Series Products Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack
RoCE	npu_chip_mac_tx_pause_num	Total Pause frames sent by MAC	Label 1	-
RoCE	npu_chip_mac_rx_pfc_pkt_num	Total PFC frames received by MAC	Label 1	-
RoCE	npu_chip_mac_tx_pfc_pkt_num	Total PFC frames sent by MAC.	Label 1	-
RoCE	npu_chip_mac_rx_bad_pkt_num	Total number of bad packets received by MAC	Label 1	-
RoCE	npu_chip_mac_tx_bad_pkt_num	Total bad packets sent by MAC	Label 1	-
RoCE	npu_chip_mac_tx_bad_oct_num	Total bytes bad packets sent by MAC	Label 1	-
RoCE	npu_chip_mac_rx_bad_oct_num	Total bytes of bad packets received by MAC	Label 1	-
RoCE	npu_chip_roce_rx_all_pkt_num	Total number of packets received by RoCE	Label 1	-
RoCE	npu_chip_roce_tx_all_pkt_num	Total number of packets sent by RoCE	Label 1	-
RoCE	npu_chip_roce_rx_err_pkt_num	Total number of bad packets received by RoCE	Label 1	-
RoCE	npu_chip_roce_tx_err_pkt_num	Total bad packets sent by RoCE	Label 1	-
RoCE	npu_chip_roce_rx_cnp_pkt_num	Number of CNP packets received by RoCE	Label 1	-
RoCE	npu_chip_roce_tx_cnp_pkt_num	Number of CNP packets sent by RoCE	Label 1	-
RoCE	npu_chip_roce_new_pkt_rty_num	Statistics on the number of transmission retries sent over RoCE	Label 1	-
RoCE	npu_chip_roce_unexpected_ack_num	Number of unexpected ACK packets received by RoCE. The NPU discards them, which does not affect services.	Label 1	-
RoCE	npu_chip_roce_out_of_order_num	Number of packets received by RoCE with a PSN greater than the expected PSN, or duplicate PSN packets. Out‑of‑order delivery or packet loss triggers transmission retries.	Label 1	-
RoCE	npu_chip_roce_verification_err_num	Number of packets received by RoCE that failed field verification. Field verification scenarios include: ICRC, packet length, destination port number, etc.	Label 1	-
RoCE	npu_chip_roce_qp_status_err_num	Number of packets received by RoCE that are generated due to abnormal QP connection status	Label 1	-
RoCE	npu_chip_info_rx_ecn_num	Number of ECN marks received by the Ascend AI Processor over the network	Label 1	-
RoCE	npu_chip_info_rx_fcs_num	Number of FCS marks by the Ascend AI Processor over the network	Label 1	-

SIO Data Information

Table 12 SIO data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
SIO	npu_chip_info_sio_crc_tx_err_cnt	Number of bad packets sent by SIO	Label 1	-	Atlas A3 Training Series Products Atlas 350 PCIe Card Atlas 850 Series Hardware Product Atlas 950 SuperPoD
SIO	npu_chip_info_sio_crc_rx_err_cnt	Number of bad packets received by SIO	Label 1	-

Optical Module Data Information

Table 13 Optical module data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
Optical Module	npu_chip_optical_state	Optical module presence status	Label 1	The value is 0 or 1. 0: Not present 1: Present	Atlas Training Series Products Atlas A2 Training Series Products Atlas 900 A3 SuperPoD Atlas 800I A2 Inference Server A200I A2 Box Heterogeneous Subrack
Optical Module	npu_chip_optical_tx_power_X (X ranges from 0 to 3)	Optical module TX power	Label 1	Unit: mW
Optical Module	npu_chip_optical_rx_power_X (X ranges from 0 to 3)	Optical module RX power	Label 1	Unit: mW
Optical Module	npu_chip_optical_vcc	Optical module voltage	Label 1	Unit: mV
Optical Module	npu_chip_optical_temp	Optical module temperature	Label 1	Unit: Celsius (°C)
Optical Module	npu_chip_info_optical_index_num_X_Y	Number of optical module lanes connected to the chip Udie Port. X is the Udie ID, and Y is the Port ID.	Label 1	-	Atlas 850 Series Hardware Product
Optical Module	npu_chip_info_optical_tx_power_Z_X_Y	Optical module TX power. Z is the lane index, with a value range of [0:3]; X is the Udie ID; Y is the Port ID.	Label 1	Unit: mW
Optical Module	npu_chip_info_optical_rx_power_Z_X_Y	Optical module RX power. Z is the lane index, with a value range of [0:3], X is the Udie ID, and Y is the Port ID.	Label 1	Unit: mW

UB Data Information

Table 14 UB data information

Category	Data Information Name	Data Information Description	Data Label	Unit	Supported Product Form
UB	npu_chip_info_ub_ipv4_pkt_cnt_rx_X_Y	Number of IPv4 UB packets received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-	Atlas 350 PCIe Card (4-Processor mesh) Atlas 850 Series Hardware Products Atlas 950 SuperPoD
UB	npu_chip_info_ub_ipv6_pkt_cnt_rx_X_Y	Number of IPv6 UB packets received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_unic_ipv4_pkt_cnt_rx_X_Y	Number of IPv4 UNIC packets received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_unic_ipv6_pkt_cnt_rx_X_Y	Number of IPv6 UNIC packets received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_ub_compact_pkt_cnt_rx_X_Y	Number of CFG6 packets received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_ub_umoc_ctph_cnt_rx_X_Y	Number of CFG7 CLAN packets received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_ub_umoc_ntph_cnt_rx_X_Y	Number of CFG7 non-CLAN packets received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_ub_mem_pkt_cnt_rx_X_Y	Number of UB mem packets received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_unknown_pkt_cnt_rx_X_Y	Number of unknown packets received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_drop_ind_cnt_rx_X_Y	Number of packets with drop_ind received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_err_ind_cnt_rx_X_Y	Number of ERR packets received on the RX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_to_host_pkt_cnt_rx_X_Y	Number of received packets on the RX side that are delivered after routing (excluding enumeration configuration and management packets). X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_to_imp_pkt_cnt_rx_X_Y	Number of enumeration configuration and management packets received on the RX side that are delivered after routing. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_to_mar_pkt_cnt_rx_X_Y	Number of UB memory packets received on the RX side that are delivered after routing. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_to_link_pkt_cnt_rx_X_Y	Number of packets received on the RX side that are forwarded to the TX side of the same port after routing. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_to_noc_pkt_cnt_rx_X_Y	Number of P2P packets received on the RX side after routing. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_route_err_cnt_rx_X_Y	Number of packets received on the RX side that encounter routing table lookup errors after routing. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_out_err_cnt_rx_X_Y	Total number of bad packets received on the RX side after verification. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_length_err_cnt_rx_X_Y	Number of packets received on the RX side that are identified as length‑error packets after validation. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_rx_busi_flit_num_X_Y	Number of RX packet bytes. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_rx_send_ack_flit_X_Y	Number of bytes in response packets returned by the RX side to the peer. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_ub_ipv4_pkt_cnt_tx_X_Y	Number of IPv4 UB packets sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_ub_ipv6_pkt_cnt_tx_X_Y	Number of IPv6 UB packets sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_unic_ipv4_pkt_cnt_tx_X_Y	Number of IPv4 UNIC packets sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_unic_ipv6_pkt_cnt_tx_X_Y	Number of IPv6 UNIC packets sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_ub_compact_pkt_cnt_tx_X_Y	Number of CFG6 packets sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_ub_umoc_ctph_cnt_tx_X_Y	Number of CFG7 CLAN packets sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_ub_umoc_ntph_cnt_tx_X_Y	Number of CFG7 non-CLAN packets sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_ub_mem_pkt_cnt_tx_X_Y	Number of UB mem packets sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_unknown_pkt_cnt_tx_X_Y	Number of unknown packets sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_drop_ind_cnt_tx_X_Y	Number of packets with drop_ind sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_err_ind_cnt_tx_X_Y	Number of ERR packets sent by the TX side. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_lpbk_ind_cnt_tx_X_Y	Number of packets sent by the TX side that are looped back via NL. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_out_err_cnt_tx_X_Y	Total number of bad packets sent by the TX side after verification. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_length_err_cnt_tx_X_Y	Number of bad packets sent by the TX side that are identified as length‑error packets after validation. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_tx_busi_flit_num_X_Y	Number of TX packet bytes. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_tx_recv_ack_flit_X_Y	Number of bytes in response packets received by the TX side from the peer. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_retry_req_sum_X_Y	Number of transmission retries. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_retry_ack_sum_X_Y	Number of response retires. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_crc_error_sum_X_Y	Number of CRC validation errors. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_core_mib_rxpausepkts_X_Y	Total number of RX pause frames. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_core_mib_txpausepkts_X_Y	Total number of TX pause frames. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_core_mib_rxpfcpkts_X_Y	Total number of RX PFC frames. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_core_mib_txpfcpkts_X_Y	Total number of TX PFC frames. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_core_mib_rxbadpkts_X_Y	Total number of RX bad packets. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_core_mib_txbadpkts_X_Y	Total number of TX bad packets. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_core_mib_rxbadoctets_X_Y	Total number of RX bad packet bytes. X is the Udie ID, and Y is the Port ID.	Label 1	-
UB	npu_chip_info_core_mib_txbadoctets_X_Y	Total number of TX bad packet bytes. X is the Udie ID, and Y is the Port ID.	Label 1	-

Data Label Information

The data labels used in the above tables are described as follows.

Table 15 Data label information

Label	Field and Description	Field Type
Label 1	container_name: Container name	String
	id: NPU ID	String
	model_name: Ascend AI Processor name. This field is composed of the chip name, chip type, and chip version. These three parts can be queried using the npu-smi info -t board -i NPU ID -c Chip ID command, and correspond respectively to Chip Name, Chip Type, and Chip Version in the return information. Note: The return information for Atlas A3 series products does not contain the Chip Type field.	String
	namespace: Namespace	String
	pcie_bus_info: PCIe information of the Ascend AI Processor	String
	pod_name: Pod name	String
	vdie_id: Unique identifier of the Ascend AI Processor, which can be used as the UUID of the NPU.	String
Label 2	container_name: Container name, in the format "Pod Namespace_Pod Name_Container Name". If the process runs on the host, this value is empty.	String
	container_id: Container ID	String
	id: NPU ID	String
	model_name: Ascend AI Processor name. This field is composed of the chip name, chip type, and chip version. These three parts can be queried using the npu-smi info -t board -i NPU ID -c Chip ID command, and correspond respectively to the Chip Name, Chip Type, and Chip Version in the return information. Note: The return information for Atlas A3 series products does not contain the Chip Type field.	String
	namespace: Namespace	String
	pcie_bus_info: PCIe information of the Ascend AI Processor	String
	pod_name: Pod name	String
	vdie_id: Unique identifier of the Ascend AI Processor, which can be used as the UUID of the NPU	String
	process_id: NPU Exporter is started by the root user in a privileged container, and the queried PID is the process PID on the host. When started in a host scenario, the queried PID is the process PID on the host. In other container scenarios, the queried PID is the process PID within the current container.	String
Label 3	aicore_count: vNPU core count	Float
	container_name: Container name	String
	id: NPU ID	String
	model_name: Ascend AI Processor name. This field is composed of the chip name, chip type, and chip version. These three parts can be queried using the npu-smi info -t board -i NPU ID -c Chip ID command, and correspond respectively to Chip Name, Chip Type, and Chip Version in the return information. [!NOTE] Note The return information for Atlas A3 series products does not contain the Chip Type field.	String
	namespace: Namespace	String
	pod_name: Pod name	String
	v_dev_id: Unique identifier of the vNPU	String
	is_virtual: Whether it is a virtual device	Bool
Label 4	container_name: Container name	String
	id: NPU ID	String
	model_name: Ascend AI Processor name. This field is composed of the chip name, chip type, and chip version. These three parts can be queried using the npu-smi info -t board -i NPU ID -c Chip ID command, and correspond respectively to the Chip Name, Chip Type, and Chip Version in the return information. Note: The return information for Atlas A3 series products does not include the Chip Type field.	String
	namespace: Namespace	String
	pcie_bus_info: PCIe information of the Ascend AI Processor	String
	pod_name: Pod name	String
	vdie_id: Unique identifier of the Ascend AI Processor, which can be used as the UUID of the NPU	String
	pcie_bw_type: Statistical value of PCIe bandwidth for remote writes. minPcieBw: Minimum value maxPcieBw: Maximum value avgPcieBw: Average value	String
Label 5	container_name: Container name	String
	containerName: Container name, in the format of "Pod Namespace_Pod Name_Container Name"	String
	containerID: Container ID	String
	npuID: NPU ID	String
	model_name: Ascend AI Processor name. This field is composed of the chip name, chip type, and chip version. These three parts can be queried using the npu-smi info -t board -i NPU ID -c Chip ID command, and correspond respectively to Chip Name, Chip Type, and Chip Version in the return information. Note: The return information for Atlas A3 series products does not include the Chip Type field.	String
	namespace: Namespace	String
	pcie_bus_info: PCIe information of the Ascend AI Processor	String
	pod_name: Pod name	String
	vdie_id: Unique identifier of the Ascend AI Processor, which can be used as the UUID of the NPU	String
Label 6	container_name: Container name	String
	id: NPU ID	String
	model_name: Ascend AI Processor name. This field is composed of the chip name, chip type, and chip version. These three parts can be queried using the npu-smi info -t board -i NPU ID -c Chip ID command, and correspond respectively to the Chip Name, Chip Type, and Chip Version in the return information. Note: The return information for Atlas A3 series products does not contain the Chip Type field.	String
	namespace: Namespace	String
	pcie_bus_info: PCIe information of the Ascend AI Processor	String
	pod_name: Pod name	String
	serial_number: Ascend AI Processor serial number. NA is reported when not configured.	String
	vdie_id: Unique identifier of the Ascend AI processor, which can be used as the UUID of the NPU.	String
Label 7	container_name: Container name.	String
	id: NPU ID	String
	model_name: Ascend AI Processor name. This field is composed of the chip name, chip type, and chip version. These three parts can be queried using the npu-smi info -t board -i NPU ID -c Chip ID command, and correspond respectively to Chip Name, Chip Type, and Chip Version in the return information. [!NOTE] Note The return information for Atlas A3 series products does not contain the Chip Type field.	String
	namespace: Namespace	String
	pcie_bus_info: PCIe information of the Ascend AI Processor	String
	pod_name: Pod name	String
	product_type: Ascend AI's processor product type	String
	vdie_id: Unique identifier of the Ascend AI Processor, which can be used as the UUID of the NPU	String

HDK Interface Call

NPU Exporter obtains corresponding information by calling the underlying HDK interfaces. For more details, refer to HDK Interfaces Called by NPU Exporter.xlsx. To find the HDK interface corresponding to the data information, refer to the following steps.

Log in to the Ascend Computing Documentation center, select and click the corresponding product name to enter the documentation page. For example, to search for the Atlas 800I A2 Inference Server documentation, click "Atlas 800I A2".
Find "Secondary Development" in the left navigation bar, and select the corresponding document based on the interface type.
- For DCMIs, select "API Reference" and click DCMI API Reference.
- For HCCN Tool, select "API Reference" and click Atlas A2 Center Inference and Training Hardware HCCN Tool API Reference.
In the search bar on the document homepage, directly search for the corresponding interface name or keyword to obtain relevant information about the interface.

Status Codes

Table 16 Status codes

Status Code	Meaning
200	Normal status
307	Temporary redirection
500	Internal server error