hwslowdown

Documentation

Overview

Package hwslowdown monitors NVIDIA GPU hardware clock events of all GPUs, such as HW Slowdown events.

Index

• Constants
• func New(ctx context.Context, cfg nvidia_common.Config, eventsStore events_db.Store) (components.Component, error)

Constants

const (
	// DefaultStateHWSlowdownEvaluationWindow is the window to evaluate the HW slowdown state.
	DefaultStateHWSlowdownEvaluationWindow = 10 * time.Minute

	// DefaultStateHWSlowdownEventsThresholdFrequencyPerMinute is the threshold frequency of the HW slowdown events per minute.
	// If the evaluation window is 10 minutes and for the last 10-minute, 6 events are found, the state is considered unhealthy, where the ratio is 0.6 = 6 / 10.
	// This is to avoid false positives when the HW slowdown events are rare.
	DefaultStateHWSlowdownEventsThresholdFrequencyPerMinute = 0.6
)

const (
	StateKeyHWSlowdown = "hw_slowdown"
)

Functions

func New

func New(ctx context.Context, cfg nvidia_common.Config, eventsStore events_db.Store) (components.Component, error)