metrics

Documentation

Index

• Variables

Variables

var (
	// MachineControllerFrozenDesc is a metric about MachineController's frozen status
	MachineControllerFrozenDesc = prometheus.NewDesc("mcm_machine_controller_frozen", "Frozen status of the machine controller manager.", nil, nil)

	// MachineCountDesc is a metric about machine count of the mcm manages
	MachineCountDesc = prometheus.NewDesc("mcm_machine_items_total", "Count of machines currently managed by the mcm.", nil, nil)

	// MachineCSPhase Current status phase of the Machines currently managed by the mcm.
	MachineCSPhase = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Namespace: namespace,
		Subsystem: machineSubsystem,
		Name:      "current_status_phase",
		Help:      "Current status phase of the Machines currently managed by the mcm.",
	}, []string{"name", "namespace"})

	// MachineInfo Information of the Machines currently managed by the mcm.
	MachineInfo = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Namespace: namespace,
		Subsystem: machineSubsystem,
		Name:      "info",
		Help:      "Information of the Machines currently managed by the mcm.",
	}, []string{"name", "namespace", "createdAt",
		"spec_provider_id", "spec_class_api_group", "spec_class_kind", "spec_class_name"})

	// MachineStatusCondition Information of the mcm managed Machines' status conditions
	MachineStatusCondition = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Namespace: namespace,
		Subsystem: machineSubsystem,
		Name:      "status_condition",
		Help:      "Information of the mcm managed Machines' status conditions.",
	}, []string{"name", "namespace", "condition"})
)

variables for subsystem: machine

var (
	// APIRequestCount Number of Cloud Service API requests, partitioned by provider, and service.
	APIRequestCount = prometheus.NewCounterVec(prometheus.CounterOpts{
		Namespace: namespace,
		Subsystem: cloudAPISubsystem,
		Name:      "requests_total",
		Help:      "Number of Cloud Service API requests, partitioned by provider, and service.",
	}, []string{"provider", "service"},
	)

	// APIFailedRequestCount Number of Failed Cloud Service API requests, partitioned by provider, and service.
	APIFailedRequestCount = prometheus.NewCounterVec(prometheus.CounterOpts{
		Namespace: namespace,
		Subsystem: cloudAPISubsystem,
		Name:      "requests_failed_total",
		Help:      "Number of Failed Cloud Service API requests, partitioned by provider, and service.",
	}, []string{"provider", "service"},
	)

	// APIRequestDuration records duration of all successful provider API calls.
	// This metric can be filtered by provider and service.
	APIRequestDuration = prometheus.NewHistogramVec(prometheus.HistogramOpts{
		Namespace: namespace,
		Subsystem: cloudAPISubsystem,
		Name:      "api_request_duration_seconds",
		Help:      "Time(in seconds) it takes for a provider API request to complete",
	}, []string{"provider", "service"})

	// DriverAPIRequestDuration records duration of all successful driver API calls.
	// This metric can be filtered by provider and operation.
	DriverAPIRequestDuration = prometheus.NewHistogramVec(prometheus.HistogramOpts{
		Namespace: namespace,
		Subsystem: cloudAPISubsystem,
		Name:      "driver_request_duration_seconds",
		Help:      "Total time (in seconds) taken for a driver API request to complete",
	}, []string{"provider", "operation"})

	// DriverFailedAPIRequests records number of failed driver API calls.
	// This metric can be filtered by provider, operation and error code.
	DriverFailedAPIRequests = prometheus.NewCounterVec(prometheus.CounterOpts{
		Namespace: namespace,
		Subsystem: cloudAPISubsystem,
		Name:      "driver_requests_failed_total",
		Help:      "Number of failed Driver API requests, partitioned by provider, operation and error code",
	}, []string{"provider", "operation", "error_code"})
)

variables for subsystem: cloud_api

var (
	// ScrapeFailedCounter is a Prometheus metric, which counts errors during metrics collection.
	ScrapeFailedCounter = prometheus.NewCounterVec(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   miscSubsystem,
		Name:        "scrape_failure_total",
		Help:        "Total count of scrape failures.",
		ConstLabels: map[string]string{"binary": "machine-controller-manager-provider"},
	}, []string{"kind"})
)

variables for subsystem: misc