Move slurm_resume configuration in a sub-recipe

Also factor out the creation of some directory and add ScalingStrategy parameter to slurm_resume config Signed-off-by: Nicola Sirena <[email protected]>
aws · Oct 5, 2023 · fe3c0a0 · fe3c0a0
1 parent efd3921
commit fe3c0a0
Show file tree

Hide file tree

Showing 5 changed files with 89 additions and 66 deletions.
diff --git a/cookbooks/aws-parallelcluster-slurm/kitchen.slurm-config.yml b/cookbooks/aws-parallelcluster-slurm/kitchen.slurm-config.yml
@@ -214,3 +214,23 @@ suites:
                 Uri: test-slurm-database.cluster-abcdefghijkl.us-east-1.rds.amazonaws.com:3306
                 UserName: clusteradmin
                 PasswordSecretArn: arn:aws:secretsmanager:us-east-1:123456789012:secret:TestSecret:abcdefghijkl-ABCDEF
+  - name: config_slurm_resume
+    run_list:
+      - recipe[aws-parallelcluster-tests::setup]
+      - recipe[aws-parallelcluster-slurm::config_slurm_resume]
+    verifier:
+      controls:
+        - config_slurm_resume
+    attributes:
+      cluster:
+        config:
+          Scheduling:
+            ScalingStrategy: test-strategy
+      dependencies:
+        - resource:nfs # This is required only for mock_slurm_dir that makes a fake export
+        - recipe:aws-parallelcluster-platform::directories
+        - recipe:aws-parallelcluster-platform::users
+        - recipe:aws-parallelcluster-slurm::slurm_users
+        - recipe:aws-parallelcluster-slurm::config_head_node_directories
+        - recipe:aws-parallelcluster-slurm::mock_slurm_dir
+
diff --git a/cookbooks/aws-parallelcluster-slurm/recipes/config/config_head_node.rb b/cookbooks/aws-parallelcluster-slurm/recipes/config/config_head_node.rb
@@ -24,26 +24,7 @@
   options ['no_root_squash']
 end unless on_docker?
 
-# Ensure config directory is in place
-directory "#{node['cluster']['slurm']['install_dir']}" do
-  user 'root'
-  group 'root'
-  mode '0755'
-end if redhat_on_docker? # we skip slurm setup on Docker UBI because we don't install python
-
-# Ensure config directory is in place
-directory "#{node['cluster']['slurm']['install_dir']}/etc" do
-  user 'root'
-  group 'root'
-  mode '0755'
-end
-
-# Create directory configured as StateSaveLocation
-directory '/var/spool/slurm.state' do
-  user node['cluster']['slurm']['user']
-  group node['cluster']['slurm']['group']
-  mode '0700'
-end
+include_recipe 'aws-parallelcluster-slurm::config_head_node_directories'
 
 template "#{node['cluster']['slurm']['install_dir']}/etc/slurm.conf" do
   source 'slurm/slurm.conf.erb'
@@ -59,14 +40,6 @@
   mode '0644'
 end
 
-# Copy pcluster config generator and templates
-remote_directory "#{node['cluster']['scripts_dir']}/slurm" do
-  source 'head_node_slurm/slurm'
-  mode '0755'
-  action :create
-  recursive true
-end
-
 unless on_docker?
   # Generate pcluster specific configs
   no_gpu = nvidia_installed? ? "" : "--no-gpu"
@@ -159,44 +132,7 @@
   mode '0644'
 end
 
-template "#{node['cluster']['scripts_dir']}/slurm/slurm_resume" do
-  source 'slurm/resume_program.erb'
-  owner node['cluster']['slurm']['user']
-  group node['cluster']['slurm']['group']
-  mode '0744'
-end
-
-file "/var/log/parallelcluster/slurm_resume.log" do
-  owner node['cluster']['cluster_admin_user']
-  group node['cluster']['cluster_admin_group']
-  mode '0644'
-end
-
-file "/var/log/parallelcluster/slurm_resume.events" do
-  owner node['cluster']['cluster_admin_user']
-  group node['cluster']['cluster_admin_group']
-  mode '0644'
-end
-
-template "#{node['cluster']['slurm_plugin_dir']}/parallelcluster_slurm_resume.conf" do
-  source 'slurm/parallelcluster_slurm_resume.conf.erb'
-  owner node['cluster']['cluster_admin_user']
-  group node['cluster']['cluster_admin_group']
-  mode '0644'
-  variables(
-    cluster_name: node['cluster']['stack_name'],
-    region: node['cluster']['region'],
-    proxy: node['cluster']['proxy'],
-    dynamodb_table: node['cluster']['slurm_ddb_table'],
-    hosted_zone: node['cluster']['hosted_zone'],
-    dns_domain: node['cluster']['dns_domain'],
-    use_private_hostname: node['cluster']['use_private_hostname'],
-    head_node_private_ip: on_docker? ? 'local_ipv4' : node['ec2']['local_ipv4'],
-    head_node_hostname: on_docker? ? 'local_hostname' : node['ec2']['local_hostname'],
-    clustermgtd_heartbeat_file_path: "#{node['cluster']['slurm']['install_dir']}/etc/pcluster/.slurm_plugin/clustermgtd_heartbeat",
-    instance_id: on_docker? ? 'instance_id' : node['ec2']['instance_id']
-  )
-end
+include_recipe 'aws-parallelcluster-slurm::config_slurm_resume'
 
 template "#{node['cluster']['scripts_dir']}/slurm/slurm_suspend" do
   source 'slurm/suspend_program.erb'

diff --git a/cookbooks/aws-parallelcluster-slurm/recipes/config/config_slurm_resume.rb b/cookbooks/aws-parallelcluster-slurm/recipes/config/config_slurm_resume.rb
@@ -0,0 +1,56 @@
+# frozen_string_literal: true
+
+#
+# Cookbook:: aws-parallelcluster-slurm
+# Recipe:: config_slurm_resume
+#
+# Copyright:: 2013-2023 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You may not use this file except in compliance with the
+# License. A copy of the License is located at
+#
+# http://aws.amazon.com/apache2.0/
+#
+# or in the "LICENSE.txt" file accompanying this file. This file is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES
+# OR CONDITIONS OF ANY KIND, express or implied. See the License for the specific language governing permissions and
+# limitations under the License.
+
+template "#{node['cluster']['scripts_dir']}/slurm/slurm_resume" do
+  source 'slurm/resume_program.erb'
+  owner node['cluster']['slurm']['user']
+  group node['cluster']['slurm']['group']
+  mode '0744'
+end
+
+file "/var/log/parallelcluster/slurm_resume.log" do
+  owner node['cluster']['cluster_admin_user']
+  group node['cluster']['cluster_admin_group']
+  mode '0644'
+end
+
+file "/var/log/parallelcluster/slurm_resume.events" do
+  owner node['cluster']['cluster_admin_user']
+  group node['cluster']['cluster_admin_group']
+  mode '0644'
+end
+
+template "#{node['cluster']['slurm_plugin_dir']}/parallelcluster_slurm_resume.conf" do
+  source 'slurm/parallelcluster_slurm_resume.conf.erb'
+  owner node['cluster']['cluster_admin_user']
+  group node['cluster']['cluster_admin_group']
+  mode '0644'
+  variables(
+    cluster_name: node['cluster']['stack_name'],
+    region: node['cluster']['region'],
+    proxy: node['cluster']['proxy'],
+    dynamodb_table: node['cluster']['slurm_ddb_table'],
+    hosted_zone: node['cluster']['hosted_zone'],
+    dns_domain: node['cluster']['dns_domain'],
+    use_private_hostname: node['cluster']['use_private_hostname'],
+    head_node_private_ip: on_docker? ? 'local_ipv4' : node['ec2']['local_ipv4'],
+    head_node_hostname: on_docker? ? 'local_hostname' : node['ec2']['local_hostname'],
+    clustermgtd_heartbeat_file_path: "#{node['cluster']['slurm']['install_dir']}/etc/pcluster/.slurm_plugin/clustermgtd_heartbeat",
+    instance_id: on_docker? ? 'instance_id' : node['ec2']['instance_id'],
+    scaling_strategy: node['cluster']['config'].dig(:Scheduling, :ScalingStrategy)
+  )
+end
diff --git a/cookbooks/aws-parallelcluster-slurm/recipes/test/mock_slurm_dir.rb b/cookbooks/aws-parallelcluster-slurm/recipes/test/mock_slurm_dir.rb
@@ -1,6 +1,7 @@
 return if on_docker?
 
 slurm_install_dir = '/opt/slurm'
+slurm_plugin_dir = '/etc/parallelcluster/slurm_plugin'
 directory slurm_install_dir do
   mode '1777'
 end
@@ -10,3 +11,12 @@
   writeable true
   options ['no_root_squash']
 end
+
+# Ensure slurm plugin directory is in place for tests that require it
+directory slurm_plugin_dir do
+  user 'root'
+  group 'root'
+  mode '0755'
+  action :create
+  recursive true
+end
diff --git a/...s/aws-parallelcluster-slurm/templates/default/slurm/parallelcluster_slurm_resume.conf.erb b/...s/aws-parallelcluster-slurm/templates/default/slurm/parallelcluster_slurm_resume.conf.erb
@@ -10,3 +10,4 @@ head_node_private_ip = <%= @head_node_private_ip %>
 head_node_hostname = <%= @head_node_hostname %>
 clustermgtd_heartbeat_file_path = <%= @clustermgtd_heartbeat_file_path %>
 instance_id = <%= @instance_id %>
+scaling_strategy = <%= @scaling_strategy %>