Installing Furiosa Metrics Exporter

Installing Furiosa Metrics Exporter#

Furiosa Metrics Exporter#

The Furiosa metrics exporter exposes collection of metrics related to FuriosaAI NPU devices in Prometheus format. In a Kubernetes cluster, you can scrape the metrics provided by furiosa-metrics-exporter using Prometheus and visualize them with a Grafana dashboard. This can be easily set up using the Prometheus Chart and Grafana Helm charts, along with the furiosa-metrics-exporter Helm chart.

Deploying Furiosa Metrics Exporter with Helm#

The Furiosa metrics exporter helm chart is available at furiosa-ai/helm-charts. To configure deployment as you need, you can modify charts/furiosa-metrics-exporter/values.yaml. For example, the Furiosa metrics exporter Helm chart automatically creates a Service Object with Prometheus annotations to enable metric scraping automatically. You can modify the values.yaml to change the port or disable the Prometheus annotations if needed. You can deploy the Furiosa Metrics Exporter by running the following commands:

helm repo add furiosa https://furiosa-ai.github.io/helm-charts
helm repo update
helm install furiosa-metrics-exporter furiosa/furiosa-metrics-exporter -n kube-system

Alternative Installation: furiosa-metrics-exporter via apt#

Note

For special use cases or non-Kubernetes environments, you can alternatively install the exporter via an apt package.

The minimum requirements for installing the furiosa-metrics-exporter package are as follows:

Ubuntu 22.04 LTS (Debian Bookworm) or later
Linux Kernel 6.3 or later
Administrator privileges on the system (root)
Setting up APT and Installing Prerequisites

Then, please install the furiosa-metrics-exporter package as follows:

sudo apt update
sudo apt install -y furiosa-metrics-exporter

This command installs packages furiosa-libsmi and furiosa-metrics-exporter.

Configuration#

The Furiosa Metrics Exporter has various command-line (CLI) options. The following table summarizes the available CLI options:

Furiosa Metrics Exporter CLI Options#
CLI Option	Required	Description
`--interval <int>`	Yes	Metrics collection interval in seconds. e.g., `--interval 10` for collecting metrics every 10 seconds.
`--port <int>`	Yes	Port number for the metrics HTTP server. e.g., `--port 6254`.
`--node-name <string>`	No	If set, hostname label will be added to the metrics. Default is empty (`""`).
`--kube-resources-label <bool>`	No	If set, Kubernetes resource labels (such as namespace, pod, and container) will be added to the metrics. Default is `false`. Recommended when running in Kubernetes for richer metrics context.

Metrics#

The exporter is composed of chain of collectors, each collector is responsible for collecting specific metrics from the Furiosa NPU devices. The following table shows the available collectors and metrics:

NPU Metrics#
Collector Name	Metric	Type	Metric Labels	Description
Liveness	furiosa_npu_alive	gauge	arch, core, device, uuid, pci_bud_id, firmware_version, pert_version, driver_version, hostname, namespace, pod, container	The liveness of the Furiosa NPU device.
Temperature	furiosa_npu_hw_temperature	gauge	arch, core, device, uuid, pci_bud_id, firmware_version, pert_version, driver_version, hostname, namespace, pod, container, label	The temperature of the Furiosa NPU device.
Power	furiosa_npu_hw_power	gauge	arch, core, device, uuid, pci_bud_id, firmware_version, pert_version, driver_version, hostname, namespace, pod, container, label	The power consumption of the Furiosa NPU device.
Core Utilization	furiosa_npu_core_utilization	gauge	arch, core, device, uuid, pci_bud_id, firmware_version, pert_version, driver_version, hostname, namespace, pod, container	The core utilization of the Furiosa NPU device.
Core Frequency	furiosa_npu_core_frequency	gauge	arch, core, device, uuid, pci_bud_id, firmware_version, pert_version, driver_version, hostname, namespace, pod, container	The core frequency of the Furiosa NPU device.
Cycle Count	furiosa_npu_total_cycle_count	counter	arch, core, device, uuid, pci_bud_id, firmware_version, pert_version, driver_version, hostname, namespace, pod, container	The total cycle count of the Furiosa NPU device.
Task Execution Cycle	furiosa_npu_task_execution_cycle	counter	arch, core, device, uuid, pci_bud_id, firmware_version, pert_version, driver_version, hostname, namespace, pod, container	The task execution cycle of the NPU Task.

All metrics share common metric labels such as arch, core, device, uuid, pci_bud_id, firmware_version, pert_version, driver_version. The following table describes the common metric labels:

Common NPU Metrics Label Attributes#
Label Attribute	Description
arch	The architecture of the Furiosa NPU device. e.g. warboy, rngd
core	The core number of the Furiosa NPU device. e.g. 0, 1, 2, 3, 4, 5, 6, 7, 0-1, 2-3, 0-3, 4-5, 6-7, 4-7, 0-7
device	The device name of the Furiosa NPU device. e.g. npu0
uuid	The UUID of the Furiosa NPU device.
pci_bus_id	The PCI bus ID of the Furiosa NPU device. e.g. 0000:c7:00.0
firmware_version	The firmware version of the Furiosa NPU device. e.g. 2025.1.0+696efad
pert_version	The pert version of the Furiosa NPU device. e.g. 2025.1.0+1694e18
driver_version	The driver version of the Furiosa NPU device. e.g. 2025.1.0+f09a8d8

The following metric labels are optional and enabled only when the corresponding command-line options (--node-name and --kube-resources-label) are set. Additionally, the namespace, pod, and container labels require an environment where the Kubernetes PodResource API is supported.

Optional NPU Metrics Label Attributes#
Label Attribute	Description
hostname	The hostname of the machine where the exporter is running.
namespace	The namespace of the Kubernetes pod to which the NPU is allocated.
pod	The name of the Kubernetes pod to which the NPU is allocated.
container	The container name within the Kubernetes pod to which the NPU is allocated.

The metric label “label” is used to describe additional attributes specific to each metric. This approach helps avoid having too many metric definitions and effectively aggregates metrics that share common characteristics.

Additional Metric Label Attributes#
Metric Type	Label Attribute	Description
Temperature	peak	The highest temperature observed from SoC sensors
Temperature	ambient	The temperature observed from sensors attached to the board
Power	rms	Root Mean Square (RMS) value of the power consumed by the device, providing an average power consumption metric over a period of time.

Examples

The following shows real-world example of the metrics:

#liveness
furiosa_npu_alive{arch="rngd",container="furiosa",core="0-7",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1

#temperature
furiosa_npu_hw_temperature{arch="rngd",container="furiosa",core="0-7",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",label="ambient",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 52
furiosa_npu_hw_temperature{arch="rngd",container="furiosa",core="0-7",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",label="peak",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 67.756

#power
furiosa_npu_hw_power{arch="rngd",container="furiosa",core="0-7",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",label="rms",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 50

#core utilization
furiosa_npu_core_utilization{arch="rngd",container="furiosa",core="0",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 99.68363645361265
furiosa_npu_core_utilization{arch="rngd",container="furiosa",core="1",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 99.68363645361265
furiosa_npu_core_utilization{arch="rngd",container="furiosa",core="2",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 99.68363645361265
furiosa_npu_core_utilization{arch="rngd",container="furiosa",core="3",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 99.68363645361265
furiosa_npu_core_utilization{arch="rngd",container="furiosa",core="4",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 99.6826341187199
furiosa_npu_core_utilization{arch="rngd",container="furiosa",core="5",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 99.6826341187199
furiosa_npu_core_utilization{arch="rngd",container="furiosa",core="6",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 99.6826341187199
furiosa_npu_core_utilization{arch="rngd",container="furiosa",core="7",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 99.6826341187199

#core frequency
furiosa_npu_core_frequency{arch="rngd",container="furiosa",core="0",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1750
furiosa_npu_core_frequency{arch="rngd",container="furiosa",core="1",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1750
furiosa_npu_core_frequency{arch="rngd",container="furiosa",core="2",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1750
furiosa_npu_core_frequency{arch="rngd",container="furiosa",core="3",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1750
furiosa_npu_core_frequency{arch="rngd",container="furiosa",core="4",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1750
furiosa_npu_core_frequency{arch="rngd",container="furiosa",core="5",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1750
furiosa_npu_core_frequency{arch="rngd",container="furiosa",core="6",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1750
furiosa_npu_core_frequency{arch="rngd",container="furiosa",core="7",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1750

#total cycle count
furiosa_npu_total_cycle_count{arch="rngd",container="furiosa",core="0",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1.7242541456e+10
furiosa_npu_total_cycle_count{arch="rngd",container="furiosa",core="1",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1.7242541456e+10
furiosa_npu_total_cycle_count{arch="rngd",container="furiosa",core="2",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1.7242541456e+10
furiosa_npu_total_cycle_count{arch="rngd",container="furiosa",core="3",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1.7242541456e+10
furiosa_npu_total_cycle_count{arch="rngd",container="furiosa",core="4",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1.7175902913e+10
furiosa_npu_total_cycle_count{arch="rngd",container="furiosa",core="5",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1.7175902913e+10
furiosa_npu_total_cycle_count{arch="rngd",container="furiosa",core="6",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1.7175902913e+10
furiosa_npu_total_cycle_count{arch="rngd",container="furiosa",core="7",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 1.7175902913e+10

#task execution cycle
furiosa_npu_task_execution_cycle{arch="rngd",container="furiosa",core="0",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 5.686392711e+09
furiosa_npu_task_execution_cycle{arch="rngd",container="furiosa",core="1",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 5.686392711e+09
furiosa_npu_task_execution_cycle{arch="rngd",container="furiosa",core="2",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 5.686392711e+09
furiosa_npu_task_execution_cycle{arch="rngd",container="furiosa",core="3",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 5.686392711e+09
furiosa_npu_task_execution_cycle{arch="rngd",container="furiosa",core="4",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 5.685170235e+09
furiosa_npu_task_execution_cycle{arch="rngd",container="furiosa",core="5",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 5.685170235e+09
furiosa_npu_task_execution_cycle{arch="rngd",container="furiosa",core="6",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 5.685170235e+09
furiosa_npu_task_execution_cycle{arch="rngd",container="furiosa",core="7",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 5.685170235e+09

When the partitioned device is allocated to a pod, device-wise metrics that are exported for each device such as liveness, temperature and power will also be exported for partitioned devices.

For example, if the device npu0 is partitioned into 2 devices (core 0-3, 4-7) and the partitioned device with core 0-3 is allocated to a pod, the temperature metrics will be exported as follows:

furiosa_npu_hw_temperature{arch="rngd",core="0-7",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",label="ambient",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",uuid="09512C86-0702-4303-8F40-474746474A40"} 50
furiosa_npu_hw_temperature{arch="rngd",core="0-7",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",label="peak",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",uuid="09512C86-0702-4303-8F40-474746474A40"} 64.41
furiosa_npu_hw_temperature{arch="rngd",container="furiosa",core="0-3",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",label="ambient",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 50
furiosa_npu_hw_temperature{arch="rngd",container="furiosa",core="0-3",device="npu0",driver_version="2025.1.0+f09a8d8",firmware_version="2025.1.0+696efad",hostname="cntk002",label="peak",namespace="default",pci_bus_id="0000:c7:00.0",pert_version="2025.1.0+1694e18",pod="furiosa",uuid="09512C86-0702-4303-8F40-474746474A40"} 64.41