This resource will help you configure auto scaling for the kubernetes cluster, see What is autoscaling config.

-> NOTE: Available since v1.127.0.

-> NOTE: From version 1.164.0, support for specifying whether to allow the scale-in of nodes by parameter scale_down_enabled.

-> NOTE: From version 1.164.0, support for selecting the policy for selecting which node pool to scale by parameter expander.

Example Usage

If you do not have an existing cluster, you need to create an ACK cluster through alicloud_cs_managed_kubernetes first, and then configure automatic scaling.

Basic Usage

variable "name" {
  default = "terraform-example"
}
data "alicloud_zones" "default" {
  available_resource_creation = "VSwitch"
}

resource "alicloud_vpc" "default" {
  vpc_name   = var.name
  cidr_block = "10.4.0.0/16"
}
resource "alicloud_vswitch" "default" {
  vswitch_name = var.name
  cidr_block   = "10.4.0.0/24"
  vpc_id       = alicloud_vpc.default.id
  zone_id      = data.alicloud_zones.default.zones.0.id
}

resource "alicloud_cs_managed_kubernetes" "default" {
  name_prefix          = var.name
  cluster_spec         = "ack.pro.small"
  worker_vswitch_ids   = [alicloud_vswitch.default.id]
  new_nat_gateway      = true
  pod_cidr             = cidrsubnet("10.0.0.0/8", 8, 36)
  service_cidr         = cidrsubnet("172.16.0.0/16", 4, 7)
  slb_internet_enabled = true
}

resource "alicloud_cs_autoscaling_config" "default" {
  cluster_id = alicloud_cs_managed_kubernetes.default.id
  // configure auto scaling
  cool_down_duration            = "10m"
  unneeded_duration             = "10m"
  utilization_threshold         = "0.5"
  gpu_utilization_threshold     = "0.5"
  scan_interval                 = "30s"
  scale_down_enabled            = true
  expander                      = "least-waste"
  skip_nodes_with_system_pods   = true
  skip_nodes_with_local_storage = false
  daemonset_eviction_for_nodes  = false
  max_graceful_termination_sec  = 14400
  min_replica_count             = 0
  recycle_node_deletion_enabled = false
  scale_up_from_zero            = true
}

Argument Reference

The following arguments are supported.

• cluster_id - (Optional, ForceNew) The id of kubernetes cluster.
• cool_down_duration - (Optional) The cool down duration. Default is 10m. If the delay (cooldown) value is set too long, there could be complaints that the Horizontal Pod Autoscaler is not responsive to workload changes. However, if the delay value is set too short, the scale of the replicas set may keep thrashing as usual.
• unneeded_duration - (Optional) The unneeded duration. Default is 10m.
• utilization_threshold - (Optional) The scale-in threshold. Default is 0.5.
• gpu_utilization_threshold - (Optional) The scale-in threshold for GPU instance. Default is 0.5.
• scan_interval - (Optional) The interval at which the cluster is reevaluated for scaling. Default is 30s.
• scale_down_enabled - (Optional) Specify whether to allow the scale-in of nodes. Default is true.
• expander - (Optional) The policy for selecting which node pool to scale. Valid values: least-waste, random, priority. For more information on these policies, see Configure auto scaling
• skip_nodes_with_system_pods - (Optional, Available since v1.209.0) If true cluster autoscaler will never delete nodes with pods from kube-system (except for DaemonSet or mirror pods). Default is true.
• skip_nodes_with_local_storage - (Optional, Available since v1.209.0) If true cluster autoscaler will never delete nodes with pods with local storage, e.g. EmptyDir or HostPath. Default is false.
• daemonset_eviction_for_nodes - (Optional, Available since v1.209.0) If true DaemonSet pods will be terminated from nodes. Default is false.
• max_graceful_termination_sec - (Optional, Available since v1.209.0) Maximum number of seconds CA waits for pod termination when trying to scale down a node. Default is 14400.
• min_replica_count - (Optional, Available since v1.209.0) Minimum number of replicas that a replica set or replication controller should have to allow their pods deletion in scale down. Default is 0.
• recycle_node_deletion_enabled - (Optional, Available since v1.209.0) Should CA delete the K8s node object when recycle node has scaled down successfully. Default is false.
• scale_up_from_zero - (Optional, Available since v1.209.0) Should CA scale up when there 0 ready nodes. Default is true.

Attributes Reference

The following attributes are exported:

• id - Resource id.

Timeouts

The timeouts block allows you to specify timeouts for certain actions:

• create - (Defaults to 90 mins) Used when creating the kubernetes cluster (until it reaches the initial running status).
• update - (Defaults to 60 mins) Used when activating the kubernetes cluster when necessary during update.
• delete - (Defaults to 60 mins) Used when terminating the kubernetes cluster.