cluster-autoscaler.md

Cluster autoscaler configuration reference for Lokomotive

Contents

• Introduction
• Prerequisites
• Configuration
• Attribute reference
• Applying
• Destroying

Introduction

The Cluster Autoscaler is a tool that automatically adjusts the size of a Kubernetes cluster when there are pods that failed to run in the cluster due to insufficient resources or when there are nodes that have been underutilized for an extended period of time and pods can be placed on other existing nodes.

Prerequisites

• A Lokomotive cluster accessible via kubectl deployed on Packet.

• For existing worker nodes in the cluster, you need to tag them manually for the Cluster Autoscaler to consider.

  k8s-cluster-${cluster_name}
  k8s-nodepool-${worker_pool}

Configuration

Currently Lokomotive supports the Cluster Autoscaler component only on the Packet platform. Support for other platforms will be added in the future.

Cluster Autoscaler component configuration example:

# cluster-autoscaler.lokocfg
component "cluster-autoscaler" {

  # Required arguments
  cluster_name = "cluster"
  worker_pool = "pool"

  # Optional arguments
  provider = "packet"
  namespace = "kube-system"
  min_workers = 1
  max_workers = 4
  scale_down_unneeded_time = "10m"
  scale_down_delay_after_add = "10m"
  scale_down_unready_time = "20m"

  packet {
    # Required arguments for Packet platform
    project_id = "0b5dc3d2-949a-447d-82cd-43fbdc1ae8c0"
    facility = "sjc1"

    # Optional arguments
    worker_type = "t1.small.x86"
    worker_channel = "stable"
  }
}

Attribute reference

Table of all the arguments accepted by the component.

Example:

Argument	Description	Default	Required
cluster_name	Name of the cluster.	-	true
worker_pool	Name of the worker pool.	-	true
namespace	Namespace where the Cluster Autoscaler will be installed.	"kube-system"	false
min_workers	Minimum number of workers in the worker pool.	1	false
max_workers	Maximum number of workers in the worker pool.	4	false
scale_down_unneeded_time	How long a node should be unneeded before it is eligible for scale down.	"10m"	false
scale_down_delay_after_add	How long scale down should wait after a scale up.	"10m"	false
scale_down_unready_time	How long an unready node should be unneeded before it is eligible for scale down.	"20m"	false
provider	Supported provider, currently Packet.	"packet"	false
packet.project_id	Packet Project ID where the cluster is running.	-	true
packet.facility	Packet Facility where the cluster is running.	-	true
packet.worker_type	Machine type for workers spawned by the Cluster Autoscaler.	"baremetal_0"	false
packet_worker_channel	Flatcar Container Linux channel to be used in workers spawned by the Cluster Autoscaler.	"stable"	false

Applying

To apply the cluster autoscaler component:

lokoctl component apply cluster-autoscaler

By default, the cluster Autoscaler pods run in the kube-system namespace

Destroying

To destroy the component:

lokoctl component render-manifest cluster-autoscaler | kubectl delete -f -