sampler

Documentation

Overview

Package sampler implements metrics-node-sampler which samples local utilization metrics from the node it is running on and serves aggregated values from a grpc API.

Samples are stored in memory over a sliding window. Responses contain values aggregated from the samples currently in memory.

Index

• type ContainerCPUThrottlingMetrics
• type ContainerKey
• type ContainerMetricType
• type Float32Slice
  • func (x Float32Slice) Len() int
  • func (x Float32Slice) Less(i, j int) bool
  • func (x Float32Slice) Swap(i, j int)
• type Int64Slice
  • func (x Int64Slice) Len() int
  • func (x Int64Slice) Less(i, j int) bool
  • func (x Int64Slice) Swap(i, j int)
• type NodeAggregatedMetricsSlice
  • func (x NodeAggregatedMetricsSlice) Len() int
  • func (x NodeAggregatedMetricsSlice) Less(i, j int) bool
  • func (x NodeAggregatedMetricsSlice) Swap(i, j int)
• type Server
  • func (s *Server) Check(ctx context.Context, _ *grpc_health_v1.HealthCheckRequest) (*grpc_health_v1.HealthCheckResponse, error)
  • func (s *Server) ListMetrics(context.Context, *api.ListMetricsRequest) (*api.ListMetricsResponse, error)
  • func (s *Server) RegisterCollectors(ctx context.Context, req *api.RegisterCollectorsRequest) (*api.RegisterCollectorsResponse, error)
  • func (s *Server) Start(ctx context.Context, stop context.CancelFunc) error
  • func (s *Server) Stop()

Types

type ContainerCPUThrottlingMetrics

type ContainerCPUThrottlingMetrics struct {
	// Time is the time that the sample was taken
	Time             time.Time
	ThrottledNanoSec uint64
	// the number of cpu scheduling periods that the container has been throttled for
	// the container has been considered throttled for a period if it attempts to use the Cpu after exceeding Cpu quota.
	ThrottledPeriods uint64
	// the total number of Cpu scheduling period that have elapsed
	TotalPeriods uint64
}

https://kernel.googlesource.com/pub/scm/linux/kernel/git/glommer/memcg/+/cpu_stat/Documentation/cgroups/cpu.txt

type ContainerKey

type ContainerKey struct {
	// ContainerID is the id of the container, and corresponds to the pod.status.containerStatuses.containerID
	ContainerID string
	// PodUID is the uid of the pod the container is running in, and corresponds to the pod.metadata.uid, or for
	// mirror pods the config.mirror annotation.
	PodUID string

	// NamespaceName is the namespace of the pod
	NamespaceName string

	// ContainerName is the name of the container
	ContainerName string

	// PodName is the name of the pod
	PodName string
}

ContainerKey is the key to a container running in a Pod

type ContainerMetricType

type ContainerMetricType string

ContainerMetricType identifies a type of metrics that corresponds to a specific cgroups file

const (
	MemoryUsageMetricType      ContainerMetricType = "memory-usage"
	MemoryOOMKillMetricTypeV1  ContainerMetricType = "oom-kill"
	MemoryOOMMetricType        ContainerMetricType = "oom"
	MemoryCurrentMetricsTypeV2 ContainerMetricType = "memory-current"
	CPUUsageMetricType         ContainerMetricType = "cpu-usage"
	CPUThrottlingMetricTypeV1  ContainerMetricType = "cpu-throttling"
)

type Float32Slice

type Float32Slice []float32

func (Float32Slice) Len

func (x Float32Slice) Len() int

func (Float32Slice) Less

func (x Float32Slice) Less(i, j int) bool

func (Float32Slice) Swap

func (x Float32Slice) Swap(i, j int)

type Int64Slice

type Int64Slice []int64

func (Int64Slice) Len

func (x Int64Slice) Len() int

func (Int64Slice) Less

func (x Int64Slice) Less(i, j int) bool

func (Int64Slice) Swap

func (x Int64Slice) Swap(i, j int)

type NodeAggregatedMetricsSlice

type NodeAggregatedMetricsSlice []*api.NodeAggregatedMetrics

func (NodeAggregatedMetricsSlice) Len

func (x NodeAggregatedMetricsSlice) Len() int

func (NodeAggregatedMetricsSlice) Less

func (x NodeAggregatedMetricsSlice) Less(i, j int) bool

func (NodeAggregatedMetricsSlice) Swap

func (x NodeAggregatedMetricsSlice) Swap(i, j int)

type Server

type Server struct {
	// MetricsNodeSampler configures the server
	samplerserverv1alpha1.MetricsNodeSampler `json:",inline" yaml:",inline"`

	// FS is the filesystem to use for reading cgroup metrics.  Defaults to the
	// OS filesystem.
	// +optional
	FS fs.FS

	// TimeFunc is the function used for getting the timestamp for when metrics
	// are read.  Defaults to time.Now.
	// +optional
	TimeFunc func(string) time.Time

	SortResults bool

	api.UnimplementedMetricsServer
	api.UnimplementedHealthServer

	CTX context.Context

	HostName string
	// contains filtered or unexported fields
}

Server runs the metrics-node-sampler server which samples cgroup metrics from the filesystem, and servers them over a GRPC API.

func (*Server) Check

func (s *Server) Check(ctx context.Context, _ *grpc_health_v1.HealthCheckRequest) (*grpc_health_v1.HealthCheckResponse, error)

func (*Server) ListMetrics

func (s *Server) ListMetrics(context.Context, *api.ListMetricsRequest) (*api.ListMetricsResponse, error)

ListMetrics lists the aggregated metrics for all containers and nodes

func (*Server) RegisterCollectors

func (s *Server) RegisterCollectors(ctx context.Context, req *api.RegisterCollectorsRequest) (*api.RegisterCollectorsResponse, error)

RegisterCollector provides an endpoint for collectors to manually register with the samplers. This is useful in cases where DNS isn't adequate. - Collector needs to get utilization data before marking itself as ready and serving metrics - DNS is slow to propagate

func (*Server) Start

func (s *Server) Start(ctx context.Context, stop context.CancelFunc) error

Start starts the server sampling metrics and serving them

func (*Server) Stop

func (s *Server) Stop()