npu-a-device-plugin serves as a management component for NPU devices in VKE clusters and supports the use of NPU devices in containers. This topic describes the common metrics of npu-a-device-plugin in the VKE cluster. To view all metrics, go to VKE monitoring metrics.

Volcengine Managed Service for Prometheus (VMP) classifies the metrics reported by VKE clusters into the following types: basic metrics, other metrics, and custom metrics. Billing methods vary based on different metric types. For more details, see Billing items.

Basic metrics

None

Other metrics

Metric name

Metric type

Metric name	Metric type
scrape_timeout_seconds	Other metrics
scrape_series_added	Other metrics
scrape_samples_scraped	Other metrics
scrape_samples_post_metric_relabeling	Other metrics
scrape_duration_seconds	Other metrics
npu_exporter_version_info	Other metrics
npu_container_info	Other metrics
npu_chip_info_voltage	Other metrics
npu_chip_info_utilization	Other metrics
npu_chip_info_used_memory	Other metrics
npu_chip_info_total_memory	Other metrics
npu_chip_info_temperature	Other metrics
npu_chip_info_process_info	Other metrics
npu_chip_info_power	Other metrics
npu_chip_info_network_status	Other metrics
npu_chip_info_name	Other metrics
npu_chip_info_link_status	Other metrics
npu_chip_info_health_status	Other metrics
npu_chip_info_hbm_used_memory	Other metrics
npu_chip_info_hbm_total_memory	Other metrics
npu_chip_info_error_code	Other metrics
npu_chip_info_bandwidth_tx	Other metrics
npu_chip_info_bandwidth_rx	Other metrics
npu_chip_info_aicore_current_freq	Other metrics
machine_npu_nums	Other metrics
container_npu_utilization	Other metrics
container_npu_used_memory	Other metrics
container_npu_total_memory	Other metrics

Last updated: 2026.08.06 10:44:02

Public security network record in Beijing: No. 11010802032137