Cost insights helps you detect stability, performance, and cost risks in ACK cluster workloads before they escalate. It tracks resource utilization across all pods and provides dedicated views for Burstable and BestEffort pods, so you can identify misconfigured workloads and act on them.
Prerequisites
Before you begin, ensure that you have:
Cost insights enabled. For more information, see Enable cost insights.
Managed Service for Prometheus enabled for your cluster. For more information, see Managed Service for Prometheus.
Why resource configuration matters
Kubernetes assigns each pod a quality of service (QoS) class based on its resource requests and limits. When a node is under resource pressure, Kubernetes uses the QoS class to decide which pods to evict first.
Guaranteed
Pods have equal resource requests and limits set for every container. These pods have the highest stability and performance and are the last to be evicted.
Burstable
Pods have a resource request but no resource limit. They can consume the full resources of the node. When eviction is needed, Kubernetes evicts BestEffort pods first, then Burstable pods.
BestEffort
Pods have no resource request or limit. They run only when the node has idle resources and are the first to be evicted when resources become insufficient.
Misconfigured resources—whether absent, too low, or too high—lead to concrete risks:
| Configuration | Resource | Risk |
|---|---|---|
| Not configured | CPU | Stability and performance are compromised; workloads may become unresponsive due to lack of CPU resources. |
| Not configured | Memory | Stability is compromised; workloads are at risk of termination due to insufficient memory. |
| Under-configured | CPU | Performance is compromised; workloads run slowly or become unresponsive. |
| Under-configured | Memory | Stability is compromised; workloads are at risk of termination due to insufficient memory. |
| Over-configured | CPU | Low resource utilization leads to unnecessary cost. |
| Over-configured | Memory | Low resource utilization leads to unnecessary cost. |
Identify risks using Stability & Efficiency Analysis
The Stability & Efficiency Analysis page, inside Cost Insights, gives you three analysis views to evaluate resource risks across the cluster.
Access the page
Log on to the ACK console. In the left-side navigation pane, click Clusters.
On the Clusters page, find the cluster you want to manage and click its name. In the left-side pane, choose Cost Suite > Cost Insights.
On the Cluster Dimension tab, go to the Stability & Efficiency Analysis page. The page shows an overview of pod counts by QoS class and their total resource usage.
View resource utilization of all pods
The Cluster pod resource usage analysis list shows basic information and resource utilization (Usage/Request) for all pods in the cluster by default. Sort or filter the list to surface workloads at the extremes:
Low utilization — the pod is over-provisioned, and you may be paying for resources that are never used.
High utilization — the pod is under-configured, which may cause performance or stability issues depending on whether the bottleneck is CPU or memory.
Utilization above 100% — usage exceeds the requested amount, which can affect workload stability.
Example: The following figure shows pods sorted by memory utilization in descending order. All displayed pods exceed 100% memory utilization and need further review to determine whether configuration adjustments are necessary.
View resource configuration of Burstable pods
The Burstable Pod-Resource Usage Analysis list shows the CPU and memory requests and limits for every Burstable pod. Because Burstable pods have no upper resource bound, missing limits can silently degrade cluster-wide performance:
No CPU limit — the pod can consume unbounded CPU, potentially starving other workloads and causing unresponsiveness.
No memory limit — the pod can exhaust node memory, putting workloads at risk of termination.
Example: The following figure shows Burstable pods without a CPU limit. These pods may degrade cluster performance or be evicted due to resource competition.
Identify unexpected BestEffort pods
The Best Effort Pod-Resource Usage Analysis list shows all BestEffort pods in the cluster. BestEffort pods carry the highest eviction risk and should be reviewed carefully. Filter and sort the list to find pods that should not be running at BestEffort quality.
Example: If critical business services are running as BestEffort pods, adjust their resource configuration to promote them to a higher QoS class.
What's next
After you identify risks with cost insights, use the following features to address them:
Resource profiling — get container-level CPU and memory recommendations based on historical usage data.
Dynamic resource overcommitment — optimize cluster cost-effectiveness by overcommitting resources based on actual usage patterns.