Define login nodes using an opentofu module (#547)

* define login nodes using tf module

* Apply suggestions from code review

Co-authored-by: Matt Anson <[email protected]>

* tweak README to explain compute groups

* try to clarify login/compute groups

---------

Co-authored-by: Matt Anson <[email protected]>

Author: sjpb

README.md

@@ -89,14 +89,19 @@ Create an OpenTofu variables file to define the required infrastructure, e.g.:
     cluster_subnet = "some_subnet" # *
     key_pair = "my_key" # *
     control_node_flavor = "some_flavor_name"
-    login_nodes = {
-        login-0: "login_flavor_name"
+    login = {
+        # Arbitrary group name for these login nodes
+        interactive = {
+            nodes: ["login-0"]
+            flavor: "login_flavor_name" # *
+        }
     }
     cluster_image_id = "rocky_linux_9_image_uuid"
     compute = {
+        # Group name used for compute node partition definition
         general = {
             nodes: ["compute-0", "compute-1"]
-            flavor: "compute_flavor_name"
+            flavor: "compute_flavor_name" # *
         }
     }
 

environments/.stackhpc/tofu/main.tf

@@ -73,8 +73,11 @@ module "cluster" {
     # are not in the same environment for stackhpc
     inventory_secrets_path = "${path.module}/../inventory/group_vars/all/secrets.yml"
 
-    login_nodes = {
-        login-0: var.other_node_flavor
+    login = {
+        login: {
+            nodes: ["login-0"]
+            flavor: var.other_node_flavor
+        }
     }
     compute = {
         standard: { # NB: can't call this default!

environments/skeleton/{{cookiecutter.environment}}/tofu/compute.tf

@@ -1,5 +1,5 @@
 module "compute" {
-  source = "./compute"
+  source = "./node_group"
 
   for_each = var.compute
 

environments/skeleton/{{cookiecutter.environment}}/tofu/nodes.tf renamed to environments/skeleton/{{cookiecutter.environment}}/tofu/control.tf

@@ -2,26 +2,6 @@ locals {
   control_volumes = concat([openstack_blockstorage_volume_v3.state], var.home_volume_size > 0 ? [openstack_blockstorage_volume_v3.home][0] : [])
 }
 
-resource "openstack_networking_port_v2" "login" {
-
-  for_each = var.login_nodes
-
-  name = "${var.cluster_name}-${each.key}"
-  network_id = data.openstack_networking_network_v2.cluster_net.id
-  admin_state_up = "true"
-
-  fixed_ip {
-    subnet_id = data.openstack_networking_subnet_v2.cluster_subnet.id
-  }
-
-  security_group_ids = [for o in data.openstack_networking_secgroup_v2.login: o.id]
-
-  binding {
-    vnic_type = var.vnic_type
-    profile = var.vnic_profile
-  }
-}
-
 resource "openstack_networking_port_v2" "control" {
 
   name = "${var.cluster_name}-control"
@@ -96,42 +76,3 @@ resource "openstack_compute_instance_v2" "control" {
   EOF
 
 }
-
-resource "openstack_compute_instance_v2" "login" {
-
-  for_each = var.login_nodes
-  
-  name = "${var.cluster_name}-${each.key}"
-  image_id = var.cluster_image_id
-  flavor_name = each.value
-  key_pair = var.key_pair
-
-  dynamic "block_device" {
-    for_each = var.volume_backed_instances ? [1]: []
-    content {
-      uuid = var.cluster_image_id
-      source_type  = "image"
-      destination_type = "volume"
-      volume_size = var.root_volume_size
-      boot_index = 0
-      delete_on_termination = true
-    }
-  }
-  
-  network {
-    port = openstack_networking_port_v2.login[each.key].id
-    access_network = true
-  }
-
-  metadata = {
-    environment_root = var.environment_root
-    k3s_token = local.k3s_token
-    control_address = [for n in openstack_compute_instance_v2.control["control"].network: n.fixed_ip_v4 if n.access_network][0]
-  }
-
-  user_data = <<-EOF
-    #cloud-config
-    fqdn: ${var.cluster_name}-${each.key}.${var.cluster_name}.${var.cluster_domain_suffix}
-  EOF
-
-}

environments/skeleton/{{cookiecutter.environment}}/tofu/inventory.tf

@@ -4,7 +4,7 @@ resource "local_file" "hosts" {
                             "cluster_name": var.cluster_name,
                             "cluster_domain_suffix": var.cluster_domain_suffix,
                             "control_instances": openstack_compute_instance_v2.control
-                            "login_instances": openstack_compute_instance_v2.login
+                            "login_groups": module.login
                             "compute_groups": module.compute
                             "state_dir": var.state_dir
                           },

environments/skeleton/{{cookiecutter.environment}}/tofu/inventory.tpl

@@ -13,12 +13,22 @@ control:
     vars:
         appliances_state_dir: ${state_dir} # NB needs to be set on group not host otherwise it is ignored in packer build!
 
-login:
+
+%{ for group_name in keys(login_groups) ~}
+${cluster_name}_${group_name}:
     hosts:
-%{ for login in login_instances ~}
-        ${ login.name }:
-            ansible_host: ${[for n in login.network: n.fixed_ip_v4 if n.access_network][0]}
-            instance_id: ${ login.id }
+%{ for node in login_groups[group_name]["compute_instances"] ~}
+        ${ node.name }:
+            ansible_host: ${node.access_ip_v4}
+            instance_id: ${ node.id }
+            image_id: ${ node.image_id }
+%{ endfor ~}
+%{ endfor ~}
+
+login:
+    children:
+%{ for group_name in keys(login_groups) ~}
+        ${cluster_name}_${group_name}:
 %{ endfor ~}
 
 %{ for group_name in keys(compute_groups) ~}

environments/skeleton/{{cookiecutter.environment}}/tofu/login.tf

@@ -0,0 +1,31 @@
+module "login" {
+  source = "./node_group"
+
+  for_each = var.login
+
+  # must be set for group:
+  nodes = each.value.nodes
+  flavor = each.value.flavor
+
+  cluster_name = var.cluster_name
+  cluster_domain_suffix = var.cluster_domain_suffix
+  cluster_net_id = data.openstack_networking_network_v2.cluster_net.id
+  cluster_subnet_id = data.openstack_networking_subnet_v2.cluster_subnet.id
+
+  # can be set for group, defaults to top-level value:
+  image_id = lookup(each.value, "image_id", var.cluster_image_id)
+  vnic_type = lookup(each.value, "vnic_type", var.vnic_type)
+  vnic_profile = lookup(each.value, "vnic_profile", var.vnic_profile)
+  volume_backed_instances = lookup(each.value, "volume_backed_instances", var.volume_backed_instances)
+  root_volume_size = lookup(each.value, "root_volume_size", var.root_volume_size)
+  extra_volumes = lookup(each.value, "extra_volumes", {})
+
+  compute_init_enable = []
+  ignore_image_changes = false
+
+  key_pair = var.key_pair
+  environment_root = var.environment_root
+  k3s_token = local.k3s_token
+  control_address = [for n in openstack_compute_instance_v2.control["control"].network: n.fixed_ip_v4 if n.access_network][0]
+  security_group_ids = [for o in data.openstack_networking_secgroup_v2.login: o.id]
+}

environments/skeleton/{{cookiecutter.environment}}/tofu/compute/main.tf renamed to environments/skeleton/{{cookiecutter.environment}}/tofu/node_group/main.tf

@@ -29,9 +29,31 @@ variable "control_node_flavor" {
     description = "Flavor name for control name"
 }
 
-variable "login_nodes" {
-  type = map
-  description = "Mapping defining login nodes: key -> (str) nodename suffix, value -> (str) flavor name"
+variable "login" {
+  type = any
+  description = <<-EOF
+    Mapping defining homogenous groups of login nodes. Multiple groups may
+    be useful for e.g. separating nodes for ssh and Open Ondemand usage, or
+    to define login nodes with different capabilities such as high-memory.
+    
+    Keys are names of groups.
+    Values are a mapping as follows:
+
+    Required:
+        nodes: List of node names
+        flavor: String flavor name
+    Optional:
+        image_id: Overrides variable cluster_image_id
+        vnic_type: Overrides variable vnic_type
+        vnic_profile: Overrides variable vnic_profile
+        volume_backed_instances: Overrides variable volume_backed_instances
+        root_volume_size: Overrides variable root_volume_size
+        extra_volumes: Mapping defining additional volumes to create and attach
+                        Keys are unique volume name.
+                        Values are a mapping with:
+                            size: Size of volume in GB
+                        **NB**: The order in /dev is not guaranteed to match the mapping
+  EOF
 }
 
 variable "cluster_image_id" {
@@ -42,8 +64,11 @@ variable "cluster_image_id" {
 variable "compute" {
     type = any
     description = <<-EOF
-        Mapping defining compute infrastructure. Keys are names of groups. Values are a
-        mapping as follows:
+        Mapping defining homogenous groups of compute nodes. Groups are used
+        in Slurm partition definitions.
+
+        Keys are names of groups.
+        Values are a mapping as follows:
 
         Required:
             nodes: List of node names