Container Service for Kubernetes:How to use cost insights for registered clusters

Prerequisites

Before you begin, make sure you have:

• Kubernetes 1.18.8 or later. If your cluster is on an earlier version, update the Kubernetes version

• Managed Service for Prometheus enabled. See Enable Managed Service for Prometheus

Limitations

• Billing data is available at 08:00:00 (UTC+8) the next day. Cost data from before you enabled cost insights is not included.

• The Day-on-day ratio and Estimated cost of next day metrics require two consecutive days of data to display.

• Weekly and monthly cumulative costs reset to zero on the first day of each billing period, because bills are generated one day after the transaction date.

• Cost analysis only tracks cloud resources exclusive to your cluster. Resources shared across multiple clusters are not included.

• If you disable the cost allocation tag (key: ack.aliyun.com) in the Expenses and Costs console, the cluster cost data stops being collected. To re-enable collection, turn on the tag with key ack.aliyun.com and the tag with key ack.alibabacloud.com/nodepool-id on the Cost allocation tags page.

• Clusters must have internet access (via a NAT gateway or equivalent) to query billing data in certain regions.

Enable cost insights

onectl addon install ack-cost-exporter

Addon ack-cost-exporter, version **** installed.

Cost definitions

Before reviewing the dashboards, understand the key cost metrics and how they are calculated:

Application costs by namespace or pod are calculated based on the bill amount. Namespaces with no CPU resource requests configured are excluded from namespace cost calculations.

Usage notes

After cost insights is enabled, billing data appears at 08:00:00 (UTC+8) the following day. Use the drop-down list in the upper-right corner of any dashboard to change the time range. The default time range is the last 7 days.

Cluster dimension

Filter criteria

Feature	Number	Description
Cluster cost analysis options	(1)	Two options: Actual billing cost with discount (default) — the payment amount for cloud resources in the cluster; Original billing cost without discount — the bill amount based on list prices. See Bill Details.
Cost allocation models	(2)	Four models: CPU model (default) — estimates pod cost by CPU resource request; Memory model — estimates pod cost by memory resource request; CPU-Memory hybrid model (recommended weights) — uses recommended CPU and memory weights; CPU-Memory hybrid model (custom weights) — uses custom weights; requires setting CPU weight setting (%). See Cost estimation policies.
Time range	(3)	The time range of data shown in the dashboard. Default is the last 7 days.

Cluster cost overview

Feature	Number	Description
Cost allocation model in use	(1)	Displays the active cost allocation model. The CPU model is selected by default.
Previous day's cost, weekly cumulative cost, and monthly cumulative cost	(2)	All figures are based on the bill amount for cloud resources in the cluster. Weekly cumulative cost resets on Monday and monthly cumulative cost resets on the first day of each month, because bills are generated one day after the transaction date. The day-on-day ratio compares the previous day's cost to the day before it. Green indicates a decrease; red indicates an increase.
Cluster cost and capacity trends	(3)	The yellow curve shows cluster cost and the blue curve shows actual cluster capacity. When the curves diverge, it typically means the average cost per CPU core has changed — check whether higher-priced resources are being used.
Real-time estimated cost per namespace	(4)	The estimated cost of each namespace, calculated as the sum of costs of all pods in that namespace.
Allocated cost per namespace	(5)	The cluster cost allocated to each namespace based on its cost ratio. Cost insights uses this formula: Σ (Pod resource request / Node capacity) × Node unit price. To get the allocated cost, multiply the total cluster cost by the namespace's cost ratio.

Stability and efficiency analysis

Feature	Number	Description
Total cluster resource usage	(1)	Shows the number of pods by QoS class and total resource usage in the cluster.
Pod resource usage	(2)	Lists basic information and resource usage (Usage/Request) for all pods. Supports filtering and sorting — use this to identify workloads with the highest or lowest resource usage.
Burstable pod resource configurations	(3)	Lists resource requests and limits for Burstable pods. Use filtering and sorting to identify resource bottlenecks and understand how much cluster resource each Burstable pod consumes.
BestEffort pod resource configurations	(4)	Lists BestEffort pods, which carry a higher stability risk due to having no resource requests or limits. Filter and sort to check for unexpected BestEffort pods and address risks early.

For details, see Use cost insights to identify risks for cluster workloads.

Cluster cost analysis

Feature	Number	Description
Cloud resource cost ratios	(1) and (2)	Shows the cost share of each cloud resource type in the cluster and how those costs change over time. Use this to identify which resource types are driving cost increases.
Total cluster cost trend	(3)	Shows daily changes in the total cluster cost.
Node pool and virtual node costs	(4)	Shows the cost of each node pool or virtual node in the cluster.
Namespace real-time estimated cost trends	(5)	Shows how estimated namespace costs change over time. Calculated using: Σ (Pod resource request / Node capacity) × Node unit price. To get the allocated cost, multiply the total cluster cost by the namespace's cost ratio.
Cluster resource request and utilization	(6)	Helps you analyze resource watermarks and identify waste. Y-axis: total allocatable resources in the cluster. Green column: resources requested within the current hour. Yellow column: actual resource usage within the current hour. Resources allocated but unused = Green column − Yellow column. Allocatable cluster resources = Y-axis − Green column. Use these metrics to: avoid wasting unallocated resources (keep ~20% unallocated as a buffer and adjust pod resource requests or downgrade instances as needed); reduce allocated-but-unused waste (identify pods with excessive requests using namespace-level stats, then adjust accordingly); right-size for periodic workloads (use watermark trends to resize cluster capacity and configure scaling policies).
Cluster billing details	(7) and (8)	Daily billing records organized by cloud resource type or instance.

Set custom pricing for on-premises nodes

By default, cost insights calculates on-premises node and application costs at CNY 0.2/vCore hour. Two configuration options let you override this default.

Option 1: Set a uniform price for all on-premises nodes

Add the DefaultIDCPricePerCPUCoreHour environment variable to the ack-cost-exporter Deployment in the kube-system namespace. The unit is CNY/vCore hour.

env:
# Set the CPU unit price to CNY 0.3/vCore hour.
- name: DefaultIDCPricePerCPUCoreHour
  value: "0.3"

Option 2: Set a price for a specific node

Add the node.kubernetes.io/price-per-day label to a node. The unit is CNY/day. The following command sets the price for a node to CNY 100/day:

kubectl label nodes <node-name> node.kubernetes.io/price-per-day="100"

FAQ

Why is no data displayed after I enable cost analysis?

Check whether you completed RAM role authorization — this is the most common cause. Monitoring data starts collecting within 3 minutes of enabling cost insights, but billing data only appears at 08:00:00 (UTC+8) the following day. Also verify that your cluster has internet access via a NAT gateway or equivalent, as some regions require it to query billing data.

Why does the cost of a namespace differ from the actual bill amount?

Namespace costs are estimates based on list prices, not actual charges. Discounts, vouchers, and cost savings plans reduce your actual cluster bill but are not reflected in the per-namespace estimate. To get a closer approximation of actual namespace cost, multiply your actual cluster bill by the namespace's cost ratio.

Why is pod cost data missing for some nodes?

The RAM role is likely missing the ecs:DescribeDisks permission. Verify the permission policy includes it. See Step 1 in the Enable cost insights topic.

Why does the bill not show all cloud services used by my cluster?

Cost analysis only collects data for cloud resources exclusively tied to your cluster, identified by the tag ack.aliyun.com:{{ClusterId}}. Resources shared across multiple clusters are not tracked. If the tag was disabled, re-enable both ack.aliyun.com and ack.alibabacloud.com/nodepool-id on the Cost allocation tags page of the Expenses and Costs console.

Why is the monthly or weekly spending lower than expected?

Cost data is only collected from the date you enabled cost insights. Historical billing data from before that date is not included in the calculations.