[SlurmTopo] Add resource for Block topology

Author: Himani Anil Deshpande

cookbooks/aws-parallelcluster-slurm/libraries/helpers.rb

@@ -242,7 +242,3 @@ def check_for_protected_mode(fleet_status_command) # rubocop:disable Lint/Nested
 def get_login_node_pool_config(config, pool_name)
   config['LoginNodes']['Pools'].select { |pool| pool['Name'] == pool_name }.first
 end
-
-def is_amazon_linux_2?
-  platform?('amazon') && node['platform_version'] == "2"
-end

cookbooks/aws-parallelcluster-slurm/libraries/update.rb

@@ -100,15 +100,3 @@ def is_login_nodes_removed?
   previous_config = YAML.safe_load(File.read(node['cluster']['previous_cluster_config_path']))
   previous_config.dig("LoginNodes") and !config.dig("LoginNodes")
 end
-
-def topology_generator_command_args
-  if node['cluster']['p6egb200_block_sizes'].nil? && are_queues_updated? && ::File.exist?("#{node['cluster']['slurm']['install_dir']}/etc/topology.conf")
-    # If topology.conf exist and Capacity Block is removed, we cleanup
-    " --cleanup"
-  elsif node['cluster']['p6egb200_block_sizes'].nil? && !are_queues_updated?
-    # We do nothing if p6e-gb200 is not used and queues are not updated
-    nil
-  else
-    " --block-sizes #{node['cluster']['p6egb200_block_sizes']}"
-  end
-end

cookbooks/aws-parallelcluster-slurm/recipes/config/config_head_node.rb

@@ -45,7 +45,7 @@
   owner 'root'
   group 'root'
   mode '0644'
-  variables( is_amazon_linux_2: is_amazon_linux_2? )
+  variables(is_amazon_linux_2: platform?('amazon') && node['platform_version'] == "2")
 end
 
 template "#{node['cluster']['slurm']['install_dir']}/etc/gres.conf" do
@@ -55,25 +55,11 @@
   mode '0644'
 end
 
-# Use slurm_parallelcluster_topology to add Block Topology plugin
-template "#{node['cluster']['slurm']['install_dir']}/etc/slurm_parallelcluster_topology.conf" do
-  source 'slurm/slurm_parallelcluster_topology.conf.erb'
-  owner 'root'
-  group 'root'
-  mode '0644'
-  not_if { is_amazon_linux_2? }
+block_topology 'Add Block Topology configuration' do
+  action :configure
 end
 
 unless on_docker?
-  # Generate Slurm topology.conf file
-  execute "generate_topology_config" do
-    command "#{cookbook_virtualenv_path}/bin/python #{node['cluster']['scripts_dir']}/slurm/pcluster_topology_generator.py"\
-              " --output-file #{node['cluster']['slurm']['install_dir']}/etc/topology.conf"\
-              " --block-sizes #{node['cluster']['p6egb200_block_sizes']}"\
-              " --input-file #{node['cluster']['cluster_config_path']}"
-    not_if { node['cluster']['p6egb200_block_sizes'].nil? || (platform?('amazon') && node['platform_version'] == "2") }
-  end
-
   # Generate pcluster specific configs
   no_gpu = nvidia_installed? ? "" : "--no-gpu"
   execute "generate_pcluster_slurm_configs" do

cookbooks/aws-parallelcluster-slurm/recipes/update/update_head_node.rb

@@ -151,22 +151,8 @@ def update_nodes_in_queue(strategy, queues)
   end
 end
 
-# Update slurm_parallelcluster_topology to add/remove Block Topology plugin
-template "#{node['cluster']['slurm']['install_dir']}/etc/slurm_parallelcluster_topology.conf" do
-  source 'slurm/slurm_parallelcluster_topology.conf.erb'
-  owner 'root'
-  group 'root'
-  mode '0644'
-  not_if { is_amazon_linux_2? }
-end
-
-# Update Slurm topology.conf file
-execute "update or cleanup topology.conf" do
-  command "#{cookbook_virtualenv_path}/bin/python #{node['cluster']['scripts_dir']}/slurm/pcluster_topology_generator.py"\
-            " --output-file #{node['cluster']['slurm']['install_dir']}/etc/topology.conf"\
-            " --input-file #{node['cluster']['cluster_config_path']}"\
-            "#{topology_generator_command_args}"
-  not_if { ::File.exist?(node['cluster']['previous_cluster_config_path']) && topology_generator_command_args.nil? || is_amazon_linux_2? }
+block_topology 'Update or Cleanup Slurm Topology' do
+  action :update
 end
 
 execute "generate_pcluster_slurm_configs" do

cookbooks/aws-parallelcluster-slurm/resources/block_topology/block_topology_alinux2023.rb

@@ -0,0 +1,19 @@
+# frozen_string_literal: true
+
+# Copyright:: 2025 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.
+
+provides :block_topology, platform: 'amazon' do |node|
+  node['platform_version'].to_i == 2023
+end
+
+use 'partial/_block_topology_common.rb'

cookbooks/aws-parallelcluster-slurm/resources/block_topology/block_topology_amazon2.rb

@@ -0,0 +1,22 @@
+# frozen_string_literal: true
+
+# Copyright:: 2025 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.
+
+provides :block_topology, platform: 'amazon', platform_version: '2'
+
+use 'partial/_block_topology_common.rb'
+
+def is_block_topology_supported?
+  # We do not support Block Topology with Alinux2 as we do not support Gb200 with this OS
+  false
+end

cookbooks/aws-parallelcluster-slurm/resources/block_topology/block_topology_redhat8.rb

@@ -0,0 +1,19 @@
+# frozen_string_literal: true
+
+# Copyright:: 2025 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.
+
+provides :block_topology, platform: 'redhat' do |node|
+  node['platform_version'].to_i >= 8
+end
+
+use 'partial/_block_topology_common.rb'

cookbooks/aws-parallelcluster-slurm/resources/block_topology/block_topology_rocky8.rb

@@ -0,0 +1,19 @@
+# frozen_string_literal: true
+
+# Copyright:: 2025 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.
+
+provides :block_topology, platform: 'rocky' do |node|
+  node['platform_version'].to_i >= 8
+end
+
+use 'partial/_block_topology_common.rb'

cookbooks/aws-parallelcluster-slurm/resources/block_topology/block_topology_ubuntu22+.rb

@@ -0,0 +1,19 @@
+# frozen_string_literal: true
+
+# Copyright:: 2025 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.
+
+provides :block_topology, platform: 'ubuntu' do |node|
+  node['platform_version'].to_i >= 22
+end
+
+use 'partial/_block_topology_common.rb'

cookbooks/aws-parallelcluster-slurm/resources/block_topology/partial/_block_topology_common.rb

@@ -0,0 +1,70 @@
+# frozen_string_literal: true
+#
+# Copyright:: 2025 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.
+
+unified_mode true
+default_action :configure
+
+action :configure do
+  return unless is_block_topology_supported?
+  # Use slurm_parallelcluster_topology to add Block Topology plugin
+  template "#{node['cluster']['slurm']['install_dir']}/etc/slurm_parallelcluster_topology.conf" do
+    source 'slurm/block_topology/slurm_parallelcluster_topology.conf.erb'
+    owner 'root'
+    group 'root'
+    mode '0644'
+  end
+  # Generate Slurm topology.conf file
+  execute "generate_topology_config" do
+    command "#{cookbook_virtualenv_path}/bin/python #{node['cluster']['scripts_dir']}/slurm/pcluster_topology_generator.py"\
+              " --output-file #{node['cluster']['slurm']['install_dir']}/etc/topology.conf"\
+              " --block-sizes #{node['cluster']['p6egb200_block_sizes']}"\
+              " --input-file #{node['cluster']['cluster_config_path']}"
+    not_if { node['cluster']['p6egb200_block_sizes'].nil? }
+  end
+end
+
+action :update do
+  return unless is_block_topology_supported?
+  # Update slurm_parallelcluster_topology to add/remove Block Topology plugin
+  template "#{node['cluster']['slurm']['install_dir']}/etc/slurm_parallelcluster_topology.conf" do
+    source 'slurm/block_topology/slurm_parallelcluster_topology.conf.erb'
+    owner 'root'
+    group 'root'
+    mode '0644'
+  end
+  # Update Slurm topology.conf file
+  execute "update or cleanup topology.conf" do
+    command "#{cookbook_virtualenv_path}/bin/python #{node['cluster']['scripts_dir']}/slurm/pcluster_topology_generator.py"\
+              " --output-file #{node['cluster']['slurm']['install_dir']}/etc/topology.conf"\
+              " --input-file #{node['cluster']['cluster_config_path']}"\
+              "#{topology_generator_command_args}"
+    not_if { ::File.exist?(node['cluster']['previous_cluster_config_path']) && topology_generator_command_args.nil? }
+  end
+end
+
+def is_block_topology_supported?
+  true
+end
+
+def topology_generator_command_args
+  if node['cluster']['p6egb200_block_sizes'].nil? && are_queues_updated? && ::File.exist?("#{node['cluster']['slurm']['install_dir']}/etc/topology.conf")
+    # If topology.conf exist and Capacity Block is removed, we cleanup
+    " --cleanup"
+  elsif node['cluster']['p6egb200_block_sizes'].nil? && !are_queues_updated?
+    # We do nothing if p6e-gb200 is not used and queues are not updated
+    nil
+  else
+    " --block-sizes #{node['cluster']['p6egb200_block_sizes']}"
+  end
+end