Kubecost Metrics

Kubecost Cost Model

The Cost Model both exports and consumes the following metrics.

Metric	Description
node_cpu_hourly_cost	Hourly cost per vCPU on this node
node_gpu_hourly_cost	Hourly cost per GPU on this node
node_ram_hourly_cost	Hourly cost per Gb of memory on this node
node_total_hourly_cost	Total node cost per hour
kubecost_load_balancer_cost	Hourly cost of a load balancer
kubecost_cluster_management_cost	Hourly cost paid as a cluster management fee
pv_hourly_cost	Hourly cost per Gb on a persistent volume
node_gpu_count	Number of GPUs available on node
container_cpu_allocation	Average number of CPUs requested/used over last 1m
container_gpu_allocation	Average number of GPUs requested over last 1m
container_memory_allocation_bytes	Average bytes of RAM requested/used over last 1m
pod_pvc_allocation	Bytes provisioned for a PVC attached to a pod
kubecost_node_is_spot	Cloud provider info about node preemptibility
kubecost_network_zone_egress_cost	Total cost per GB egress across zones
kubecost_network_region_egress_cost	Total cost per GB egress across regions
kubecost_network_internet_egress_cost	Total cost per GB of internet egress
service_selector_labels	Service Selector Labels
deployment_match_labels	Deployment Match Labels
statefulSet_match_labels	StatefulSet Match Labels
kubecost_cluster_memory_working_set_bytes	(Created by recording rule)

Kubecost Network Costs

The Kubecost network-costs DaemonSet collects node network data and exports the egress, ingress, and performance statistics.

Metric	Description
kubecost_pod_network_egress_bytes_total	egressed byte counts by pod
kubecost_pod_network_ingress_bytes_total	ingressed byte counts by pod
kubecost_network_costs_parsed_entries	total parsed conntrack entries
kubecost_network_costs_parse_time	total time in milliseconds it took to parse conntrack entries

cAdvisor

cAdvisor (Container Advisor) provides container users an understanding of the resource usage and performance characteristics of their running containers. It is a running daemon that collects, aggregates, processes, and exports information about running containers.

GitHub: https://github.com/google/cadvisor

Metric	Description
container_memory_usage_bytes	Current memory usage, including all memory regardless of when it was accessed
container_fs_limit_bytes	Number of bytes that can be consumed by the container on this filesystem
container_fs_usage_bytes	Number of bytes that are consumed by the container on this filesystem
container_memory_working_set_bytes	Current working set
container_network_receive_bytes_total	Cumulative count of bytes received
container_network_transmit_bytes_total	Cumulative count of bytes transmitted
container_cpu_usage_seconds_total	Cumulative cpu time consumed
container_cpu_cfs_periods_total	Number of elapsed enforcement period intervals
container_cpu_cfs_throttled_periods_total	Number of throttled period intervals

Kube-State-Metrics (KSM)

Although the default Kubecost installation does not include a KSM deployment, Kubecost does calculate & emit the below metrics. The below metrics and labels follow conventions of KSMv1, not KSMv2.

Metric	Description
kube_deployment_spec_replicas	Number of pods specified for a Deployment
kube_deployment_status_replicas_available	Number of pods currently available for a Deployment
kube_job_status_failed	The number of pods which reached Phase Failed and the reason for failure
kube_namespace_annotations	Kubernetes annotations converted to Prometheus labels
kube_namespace_labels	Kubernetes labels converted to Prometheus labels
kube_node_labels	Kubernetes labels converted to Prometheus labels
kube_node_status_allocatable	The allocatable for different resources of a node that are available for scheduling
kube_node_status_allocatable_cpu_cores	Total allocatable cpu cores of the node (Deprecated in ksm 2.0.0)
kube_node_status_allocatable_memory_bytes	Total allocatable memory bytes of the node (Deprecated in ksm 2.0.0)
kube_node_status_capacity	The capacity for different resources of a node
kube_node_status_capacity_cpu_cores	Total cpu cores available on the the node (Deprecated in ksm 2.0.0)
kube_node_status_capacity_memory_bytes	Total memory available on the node (bytes) (Deprecated in ksm 2.0.0)
kube_node_status_condition	The condition of a cluster node
kube_persistentvolume_capacity_bytes	Total capacity of a persistent volume (bytes)
kube_persistentvolume_status_phase	Status of a persistent volume (Bound
kube_persistentvolumeclaim_info	Information about persistent volume claim
kube_persistentvolumeclaim_resource_requests_storage_bytes	The capacity of storage requested by the persistent volume claim
kube_pod_annotations	Kubernetes annotations converted to Prometheus labels
kube_pod_container_resource_limits	The number of requested limit resource by a container
kube_pod_container_resource_limits_cpu_cores	Limit on CPU cores that can be used by the container. (Deprecated in ksm 2.0.0)
kube_pod_container_resource_limits_memory_bytes	Limit on the amount of memory that can be used by the container. (Deprecated in ksm 2.0.0)
kube_pod_container_resource_requests	The number of requested request resource by a container
kube_pod_container_status_restarts_total	The number of container restarts per container
kube_pod_container_status_running	Describes whether the container is currently in running state
kube_pod_container_status_terminated_reason	Describes the reason the container is currently in terminated state
kube_pod_labels	Kubernetes labels converted to Prometheus labels
kube_pod_owner	Information about the Pod's owner
kube_pod_status_phase	The pods current phase (Pending
kube_replicaset_owner	Information about the ReplicaSet's owner

Node exporter

Prometheus exporter for hardware and OS metrics exposed by *NIX kernels, written in Go with pluggable metric collectors.

The Node Exporter is disabled by default. You can enable it with the flags:

--set prometheus.server.nodeExporter.enabled=true
--set prometheus.serviceAccounts.nodeExporter.create=true

GitHub: https://github.com/prometheus/node_exporter

Metric	Description
node_cpu_seconds_total	Seconds the cpus spent in each mode
node_disk_reads_completed	The total number of reads completed successfully
node_disk_reads_completed_total	The total number of reads completed successfully
node_disk_writes_completed	The total number of writes completed successfully
node_disk_writes_completed_total	The total number of writes completed successfully
node_filesystem_device_error	Whether an error occurred while getting statistics for the given device
node_memory_Buffers_bytes	Memory information field Buffers_bytes
node_memory_Cached_bytes	Memory information field Cached_bytes
node_memory_MemAvailable_bytes	Memory information field MemAvailable_bytes
node_memory_MemFree_bytes	Memory information field MemFree_bytes
node_memory_MemTotal_bytes	Memory information field MemTotal_bytes
node_network_transmit_bytes_total	Network device statistic transmit_bytes

Prometheus

Prometheus emits metrics which are used by Kubecost for diagnostic purposes:

Metric	Description
up	Scrape target status
prometheus_target_interval_length_seconds	Amount of time between target scrapes

NVIDIA GPUs

NVIDIA GPU monitoring support can be explained in more detail in the Kubecost Docs: NVIDIA GPU Monitoring Configurations and on the Kubecost Blog: Monitoring NVIDIA GPU Usage in Kubernetes with Prometheus. Monitoring of NVIDIA GPUs requires DCGM Exporter. While all metrics exposed by DCGM Exporter are collected, the following metrics are the ones currently used by Kubecost:

Metric	Description
DCGM_FI_PROF_GR_ENGINE_ACTIVE	GPU utilization