CCE gpu node pool Module

A module designed to create a cce gpu node pool.

Usage example:

# First you need to create a normal cce cluster
# https://github.com/iits-consulting/terraform-opentelekomcloud-project-factory/tree/master/modules/cce
module "cce" {
  source = "iits-consulting/cce-gpu-node-pool/opentelekomcloud"
  name   = "mycompany-dev-cluster"

  cluster_vpc_id            = module.vpc.vpc.id
  cluster_subnet_id         = values(module.vpc.subnets)[0].id
  cluster_high_availability = false
  cluster_enable_scaling    = false
  node_availability_zones   = ["eu-de-03"]
  node_count                = 3
  node_flavor               = "s3.large.8"
  node_storage_type         = "SSD"
  node_storage_size         = 100
}

module "cce_gpu_node_pool" {
  source = "iits-consulting/cce-gpu-node-pool/opentelekomcloud"

  name_prefix                     = module.cce.cluster_name
  cce_cluster_id                  = module.cce.cluster_id
  node_availability_zones         = ["eu-de-03"]
  node_flavor                     = "pi2.2xlarge.4"
  node_storage_type               = "SSD"
  node_storage_size               = 100
  node_count                      = 1
  autoscaler_node_max             = 1
  gpu_driver_url                  = "https://us.download.nvidia.com/tesla/535.129.03/NVIDIA-Linux-x86_64-535.129.03.run"
}

Notes

• containernetwork_type of the cce needs to be _vpc-router

Requirements

Name	Version
terraform	>= 1.5.7
errorcheck	3.0.3
opentelekomcloud	>= 1.36.34, ~>1.36
random	~> 3.0
tls	~> 4.0

Providers

Name	Version
errorcheck	3.0.3
opentelekomcloud	>= 1.36.34, ~>1.36
random	~> 3.0
tls	~> 4.0

Modules

No modules.

Resources

Name	Type
errorcheck_is_valid.gpu_beta_version_availability	resource
errorcheck_is_valid.gpu_driver_url	resource
errorcheck_is_valid.node_availability_zones	resource
opentelekomcloud_cce_addon_v3.gpu	resource
opentelekomcloud_cce_node_pool_v3.cluster_node_pool	resource
opentelekomcloud_compute_keypair_v2.cluster_keypair	resource
opentelekomcloud_kms_key_v1.node_storage_encryption_key	resource
random_id.cluster_keypair_id	resource
random_id.id	resource
tls_private_key.cluster_keypair	resource
opentelekomcloud_cce_addon_templates_v3.gpu	data source
opentelekomcloud_cce_cluster_v3.cluster	data source
opentelekomcloud_identity_project_v3.current	data source
opentelekomcloud_kms_key_v1.node_storage_encryption_existing_key	data source

Inputs

Name	Description	Type	Default	Required
cce_cluster_id	ID of the existing CCE cluster.	string	n/a	yes
name_prefix	Name prefix for provisioned resources.	string	n/a	yes
node_availability_zones	Availability zones for the node pools. Providing multiple availability zones creates one node pool in each zone.	set(string)	n/a	yes
node_count	Number of nodes to create	number	n/a	yes
node_flavor	Node specifications in otc flavor format	string	n/a	yes
autoscaler_node_max	Maximum limit of servers to create. (default: 10)	number	10	no
autoscaler_node_min	Lower bound of servers to always keep (default: <node_count>)	number	null	no
gpu_beta_enabled	Install GPU Beta Addon	bool	true	no
gpu_beta_version	Version of the GPU Beta Addon Template (CCE AI Suite)	string	"latest"	no
gpu_driver_url	Nvidia Driver download URL. Please refer to https://www.nvidia.com/Download/Find.aspx and ensure your driver is matching the GPU in your node flavor.	string	""	no
node_container_runtime	The container runtime to use. Must be set to either containerd or docker. (default: containerd)	string	"containerd"	no
node_k8s_tags	(Optional, Map) Tags of a Kubernetes node, key/value pair format.	map(string)	{}	no
node_os	Operating system of worker nodes.	string	"EulerOS 2.9"	no
node_postinstall	Post install script for the node pool.	string	""	no
node_scaling_enabled	Enable the scaling for the node pool. Please note that CCE cluster must have autoscaling addon installed. (default: 10)	bool	true	no
node_storage_encryption_enabled	Enable OTC KMS volume encryption for the node pool volumes. (default: false)	bool	false	no
node_storage_encryption_kms_key_name	If KMS volume encryption is enabled, specify a name of an existing kms key. Setting this disables the creation of a new kms key. (default: null)	string	null	no
node_storage_size	Size of the node data disk in GB (default: 100)	number	100	no
node_storage_type	Type of node storage SATA, SAS or SSD (default: SATA)	string	"SATA"	no
node_taints	Enable the scaling for the node pool. Please note that CCE cluster must have autoscaling addon installed. (default: 10)	list(object({ effect = string key = string value = string }))	[ { "effect": "PreferNoSchedule", "key": "gpu-node", "value": "true" } ]	no
tags	Common tags to add to resources that support them.	map(string)	null	no

Outputs

No outputs.