Ternary Integrations: Google Cloud Kubernetes Measure Definitions

Ternary provides standardized Kubernetes measures to help you analyze cost, usage, and resource efficiency across Google Kubernetes Engine (GKE) clusters. These measures are derived from metrics collected via Google Cloud Monitoring and are used to evaluate resource requests, actual usage, and resulting waste.

Kubernetes Container Measure Definition

Metric	Definition
limitCPUCores	Core-Days
limitMemoryBytes	Byte-Days
requestCPUCores	k8s_container.request_cpu_cores / 24.0 = requestCPUCores in Core-Days
requestCPUCost	(requestCPUCores) * (CPU USD Price) = Dollars
requestMemoryBytes	k8s_container.request_memory_bytes / 24.0 = requestMemoryBytes in Byte-Days
requestMemoryCost	k8s_container.request_memory_bytes * k8s_node_detail.ram_usd_price / 1073741824.0 = requestMemoryCost in USD (There are 1073741824.0 bytes per GiB)
totalRequestCost	requestMemoryCost + requestCPUCost = totalRequestCost in USD
totalWasteCost	requestMemoryCost + requestCPUCost - usedCPUCost - usedMemoryCost = totalWasteCost in USD
usedCPUCores	k8s_container.used_cpu_cores / 24.0 = usedCPUCores in Core-Days
usedCPUCost	k8s_container.used_cpu_cores * k8s_node_detail.cpu_usd_price = usedCPUCost in USD
usedMemoryBytes	k8s_container.used_memory_bytes / 24.0 = usedMemoryBytes in Byte-Days
usedMemoryCost	k8s_container.used_memory_bytes * k8s_node_detail.ram_usd_price / 1073741824.0 = usedMemoryCost in USD
wasteCPUs	requestCPUCores - usedCPUCores = wasteCPUs in Core-Days
wasteMemoryBytes	requestMemoryBytes - usedMemoryBytes = wasteMemoryBytes in Byte-Days
wasteCPUCost	requestCPUCost - usedCPUCost = wasteCPUCost in USD
wasteMemoryCost	requestMemoryCost - usedMemoryCost = wasteMemoryCost in USD

Kubernetes Node Measure Definition

Metric	Definition
allocatableCPUUtilization	allocatable_cpu_utilization / 24.0 = allocatableCPUUtilization
allocatableMemoryUtilization	allocatable_memory_utilization / 24.0 = allocatableMemoryUtilization
CPUCoreUsageTime	cpu_core_usage_time / 24.0 = CPUCoreUsageTime in Core-Days
totalCost	totalMemoryCost + totalCPUCost = totalCost in USD
totalCPUCores	k8s_node.total_cpu_cores / 24.0 = totalCPUCores in Core-Days
totalCPUCost	k8s_node.total_cpu_cores * k8s_node_detail.cpu_usd_price = totalCPUCost in USD
totalMemoryCost	k8s_node.total_memory_bytes * k8s_node_detail.ram_usd_price / 1073741824.0 = totalMemoryCost in USD
totalMemoryBytes	k8s_node.total_memory_bytes / 24.0 = totalMemoryBytes in Byte-Days
usedMemoryBytes	used_memory_bytes / 24.0 = usedMemoryBytes in Byte-Days
wasteCPUs	totalCPUCores - CPUCoreUsagetime = wasteCPUs in Core-Days
wasteMemoryBytes	totalMemorybytes - usedMemoryBytes = wasteMemoryBytes in Bytes

Troubleshooting Missing Kubernetes Metrics

Ternary collects Kubernetes metrics using Google Cloud Monitoring, which is enabled by default for GKE clusters.

Cloud Monitoring is enabled

If Cloud Monitoring is not enabled, metrics for clusters, namespaces, workloads, and pods will not be available in Ternary.

Service account permissions

If Cloud Monitoring is enabled but metrics are still missing, the issue is typically related to permissions. This commonly occurs when a custom service account is used instead of the default Compute Engine service account.

Ensure the following roles are assigned to the node pool service account:

roles/logging.logWriter
roles/monitoring.metricWriter
roles/stackdriver.resourceMetadata.writer

Verify metrics in Google Cloud Monitoring

Confirm whether metrics are being published:

Navigate to Metrics Explorer in Google Cloud Console
Run the following Kubernetes metric query:

* Metric: Kubernetes Container - CPU Usage time
* Filters: project_id=your_project_id
* groupingFunction: Sum
* groupByLabels:
  * cluster_name
  * node_name
  * top_level_controller_name
  * top_level_controller_type
  * container_name
  * location
  * namespace_name
  * pod_name
* perSeriesAligner: Rate
* alignmentPeriod: 1h

If the query returns no results, Kubernetes metrics are not being published to Cloud Monitoring.