Node Consolidation

Available in v1.20.6+

The Node Consolidation feature allows to instantly trigger node consolidation in EKS, AKS and GKE clusters, as well as optimize Cluster Autoscaler configurations. This feature is designed to accelerate cost savings and resource efficiency by consolidation of underutilized nodes without waiting for standard Cluster Autoscaler or Karpenter operations or manual intervention.

Node Consolidation leverages Cluster Autoscaler’s and Karpenter’s core capabilities, while adding intelligent consolidation, safety checks, and optional instance type replacement logic to improve both efficiency and reliability.

Key Features

In order to get started with Node Consolidation, integrate your AWS, Azure, or Google Cloud account by completing the Cloud Integration. This integration provides the necessary authentication and permissions for ScaleOps to manage your cluster nodes.

Node Consolidation

The Node Consolidation page captures the potential savings across the cluster, and allows to execute node consolidation. The Node Consolidation table presents the Current state of node types, and shows the recommended state post consolidation.

When clicking Consolidate Now, ScaleOps will initiate a consolidation process, according to the Consolidation Settings:

• Scale down nodes with unevictable pods: When enabled, consolidation will scale down nodes with unevictable pods (pods with annotations, pods of stateful workloads, etc)
• Scale down nodes with min count: When enabled, consolidation will scale down nodes from node pools having the minimum amount of nodes.
• Scale down nodes with local storage: When enabled, consolidation will scale down nodes with local storage.
• Scale down nodes with kube-system pods: When enabled, consolidation will scale down nodes with kube-system pods.
• Replace nodes with better instance types: When enabled, during consolidation ScaleOps will also replace nodes with a cheaper instance type still capable of accommodating the pods running on the node. Supports both Cluster Autoscaler–managed and Karpenter-managed nodes.
• Scale down Cluster Autoscaler nodes: When enabled on AWS clusters that run both Karpenter and in-cluster Cluster Autoscaler, consolidation recommendations include Cluster Autoscaler–managed nodes as well as Karpenter-managed nodes. Available in v1.29.12+

Consolidation settings and schedule can also be configured via Helm—see Node Consolidation.

Schedule Consolidation Available in v1.24.2+

Node Consolidation can be scheduled to run automatically at predefined times. This allows you to automate consolidation during low-traffic periods without triggering it manually.

Configuration

Click Schedule consolidation, and select times to automatically trigger consolidations.

You’ll see the next scheduled run in the Node Consolidation page. All times are in local timezone.

Behavior

When a scheduled time arrives:

• The consolidation will execute with the current applied consolidation settings (at the bottom of the page)
• If no candidate nodes exist (cluster already consolidated), the run is skipped.
• If a consolidation (manual or scheduled) is already in progress, a new one will not be triggered.

Mixed Karpenter and Cluster Autoscaler on AWS

Available in v1.29.12+

Mixed-cluster support extends consolidation to clusters that run both cluster autoscalers: you can get consolidation recommendations for Karpenter-managed nodes and Cluster Autoscaler–managed nodes in the same cluster, and tune Cluster Autoscaler–specific consolidation options when relevant.

How it works

With Cloud Integration configured and consolidation enabled, turn on Scale down Cluster Autoscaler nodes in Node Consolidation consolidation settings. ScaleOps then surfaces consolidation candidates for:

• Nodes provisioned by Karpenter (subject to Karpenter-related settings), and
• Nodes backed by Cluster Autoscaler node groups (subject to Cluster Autoscaler–specific settings such as local storage, kube-system pods, min count, and instance replacement where applicable).

In Node Management, both Karpenter optimization pages and Cluster Autoscaler configuration pages remain available so you can work with each stack in one place.

Instance Type Replacement on Karpenter Nodes Available in v1.30.7+

When Replace nodes with better instance types is enabled, ScaleOps searches for a cheaper instance type within the node pool’s allowed instance types that can still accommodate the current node’s workload. If one is found, the node is cordon-drained during consolidation plan execution and Karpenter provisions the cheaper replacement node.

Behavior and constraints:

• In spot node pools, the specific instance type Karpenter provisions is subject to spot capacity availability.
• Nodes from node pools that allow both spot and on-demand capacity types are not eligible for replacement.
• Replacement respects the Scale down nodes with GPU and Scale down nodes with unevictable pods settings.
• On clusters running both Karpenter and Cluster Autoscaler: if Scale down Cluster Autoscaler nodes is enabled but the Cluster Autoscaler expander is not configured to least-waste, ScaleOps skips Cluster Autoscaler–managed nodes for replacement and displays a banner prompting you to configure the Least-Waste expander.

How to enable: Go to Node Consolidation → Node Consolidation Settings → enable Replace nodes with better instance types.

Cluster Autoscaler Configuration

The Cluster Autoscaler Configuration page is available for Cluster Autoscaler on AWS, Azure, and GCP. In this page you can review the configuration of your Cluster Autoscaler, and edit some of its configuration along with ScaleOps recommendations for optimizing it.

Prerequisites

In order to use Node Consolidation the following prerequisites are required:

• Cloud Integration enabled.

• AWS (Cluster Autoscaler only): Cluster Autoscaler installed in-cluster. Both EKS Managed Node Groups and Self-Managed Autoscaling Groups are supported. The Cluster Autoscaler expander needs to be set to least-waste in order to support node pool replacement during consolidation.

• AWS (Karpenter only): Karpenter installed; Cluster Autoscaler-specific prerequisites above do not apply.

• AWS (Karpenter and Cluster Autoscaler together): Both installed in-cluster; enable Scale down Cluster Autoscaler nodes in consolidation settings to include Cluster Autoscaler–managed nodes. Available in v1.29.12+

• AKS / GKE: Cloud Managed Cluster Autoscaler. Self-Managed Cluster Autoscaler is not supported.

• AKS: The Cluster Autoscaler expander needs to be set to least-waste in order to support node pool replacement during consolidation.

Node Pool Modification Behavior

ScaleOps will edit/create node pools in some consolidation cases, as described below:

• When “Replace nodes with better instance types” is enabled on Cluster Autoscaler–managed nodes, ScaleOps creates new node pools so the Cluster Autoscaler can select optimal instance types.

  • When a node is identified for replacement with a more suitable instance type, ScaleOps duplicates the existing node pool configuration (including taints) and updates only the instance type
  • ScaleOps only configures new node pools without modifying your existing node pools
  • This ensures that newly created nodes retain the same taints as the original nodes, maintaining your workload placement constraints

• When “Replace nodes with better instance types” is enabled on Karpenter-managed nodes, no new node pools are created. ScaleOps cordon-drains the node and Karpenter provisions the replacement node from the node pool’s allowed instance types.

• When “Scale down node with min count” is enabled, ScaleOps reduces the node pool minimum if a scale-down candidate is blocked by the min count.