nvidia

Documentation

Overview

Package nvidia holds the logic to collect metrics from the NVIDIA Management Library (NVML). The main entry point is the BuildCollectors functions, which returns a set of collectors that will gather metrics from the available NVIDIA devices on the system. Each collector is responsible for a specific subsystem of metrics, such as device metrics, GPM metrics, etc. The collected metrics will be returned with the associated tags for each device.

Index

• type Collector
  • func BuildCollectors(lib nvml.Interface) ([]Collector, error)
• type Metric

Types

type Collector

type Collector interface {
	// Collect collects metrics from the given NVML device. This method should not fill the tags
	// unless they're metric-specific (i.e., all device-specific tags will be added by the Collector itself)
	Collect() ([]Metric, error)

	// Close closes the subsystem and releases any resources it might have allocated
	Close() error

	// name returns the name of the subsystem
	Name() string
}

Collector defines a collector that gets metric from a specific NVML subsystem and device

func BuildCollectors

func BuildCollectors(lib nvml.Interface) ([]Collector, error)

BuildCollectors returns a set of collectors that can be used to collect metrics from NVML.

type Metric

type Metric struct {
	Name  string   // Name holds the name of the metric.
	Value float64  // Value holds the value of the metric.
	Tags  []string // Tags holds the tags associated with the metric.
}

Metric represents a single metric collected from the NVML library.