Refactor environment recipes to be clearer in functional description …

…and remove unnecessary recipes
aws · Sep 20, 2023 · 1364a59 · 1364a59
1 parent 60d9074
commit 1364a59
Show file tree

Hide file tree

Showing 18 changed files with 122 additions and 234 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -10,6 +10,7 @@ This file is used to list changes made in each version of the AWS ParallelCluste
 - Install [Spack](https://spack.io) by default in cluster user's home directory.
 
 **CHANGES**
+- Migrate NFS exports from the head node root volume and associated data for intra-cluster shared storage to external AWS EFS filesystems attached to the cluster stack.
 
 **BUG FIXES**
 - Fix inconsistent scaling configuration after cluster update rollback when modifying the list of instance types declared in the Compute Resources.

diff --git a/cookbooks/aws-parallelcluster-entrypoints/recipes/update.rb b/cookbooks/aws-parallelcluster-entrypoints/recipes/update.rb
@@ -17,7 +17,7 @@
 end
 
 # generate the updated shared storages mapping file
-include_recipe 'aws-parallelcluster-environment::fs_update'
+include_recipe 'aws-parallelcluster-environment::update_fs_mapping'
 
 include_recipe 'aws-parallelcluster-environment::directory_service'
 include_recipe 'aws-parallelcluster-slurm::update' if node['cluster']['scheduler'] == 'slurm'

diff --git a/cookbooks/aws-parallelcluster-environment/kitchen.environment-config.yml b/cookbooks/aws-parallelcluster-environment/kitchen.environment-config.yml
@@ -414,10 +414,10 @@ suites:
         fsx_shared_dirs: ''
         raid_shared_dir: ''
         ephemeral_dir: test1
-  - name: fs_update
+  - name: update_fs_mapping
     run_list:
       - recipe[aws-parallelcluster-tests::setup]
-      - recipe[aws-parallelcluster-environment::fs_update]
+      - recipe[aws-parallelcluster-environment::update_fs_mapping]
     verifier:
       controls:
         - fs_data_file_created_correctly
@@ -440,10 +440,10 @@ suites:
         fsx_dns_names: dns1,dns2
         fsx_mount_names: mount1,mount2
         fsx_volume_junction_paths: value1,value2
-  - name: fs_update_default_values
+  - name: update_fs_mapping_default_values
     run_list:
       - recipe[aws-parallelcluster-tests::setup]
-      - recipe[aws-parallelcluster-environment::fs_update]
+      - recipe[aws-parallelcluster-environment::update_fs_mapping]
     verifier:
       controls:
         - fs_data_file_with_default_values
@@ -472,42 +472,43 @@ suites:
         scheduler: slurm
         head_node_imds_secured: 'true'
         head_node_imds_allowed_users: ['root', 'nobody']
-  - name: mount_shared_compute
-    run_list:
-      - recipe[aws-parallelcluster-tests::setup]
-      - recipe[aws-parallelcluster-environment::mount_shared]
-    verifier:
-      controls:
-        - mount_home
-        - mount_shared_compute
-    attributes:
-      dependencies:
-        - recipe:aws-parallelcluster-platform::directories
-        - resource:nfs
-        - recipe:aws-parallelcluster-environment::mock_export_directories
-      cluster:
-        node_type: 'ComputeFleet'
-        head_node_private_ip: '127.0.0.1'
-        head_node_home_path: '/fake_headnode_home'
-        shared_dir_head: '/fake_headnode_shared'
-  - name: mount_shared_login
-    run_list:
-      - recipe[aws-parallelcluster-tests::setup]
-      - recipe[aws-parallelcluster-environment::mount_shared]
-    verifier:
-      controls:
-        - mount_home
-        - mount_shared_login
-    attributes:
-      dependencies:
-        - recipe:aws-parallelcluster-platform::directories
-        - resource:nfs
-        - recipe:aws-parallelcluster-environment::mock_export_directories
-      cluster:
-        node_type: 'LoginNode'
-        head_node_private_ip: '127.0.0.1'
-        head_node_home_path: '/fake_headnode_home'
-        shared_dir_head: '/fake_headnode_shared'
+# TODO replacement for the mount_internal_use_fs recipe since it uses shared storage
+#  - name: mount_shared_compute
+#    run_list:
+#      - recipe[aws-parallelcluster-tests::setup]
+#      - recipe[aws-parallelcluster-environment::mount_shared]
+#    verifier:
+#      controls:
+#        - mount_home
+#        - mount_shared_compute
+#    attributes:
+#      dependencies:
+#        - recipe:aws-parallelcluster-platform::directories
+#        - resource:nfs
+#        - recipe:aws-parallelcluster-environment::mock_export_directories
+#      cluster:
+#        node_type: 'ComputeFleet'
+#        head_node_private_ip: '127.0.0.1'
+#        head_node_home_path: '/fake_headnode_home'
+#        shared_dir_head: '/fake_headnode_shared'
+#  - name: mount_shared_login
+#    run_list:
+#      - recipe[aws-parallelcluster-tests::setup]
+#      - recipe[aws-parallelcluster-environment::mount_shared]
+#    verifier:
+#      controls:
+#        - mount_home
+#        - mount_shared_login
+#    attributes:
+#      dependencies:
+#        - recipe:aws-parallelcluster-platform::directories
+#        - resource:nfs
+#        - recipe:aws-parallelcluster-environment::mock_export_directories
+#      cluster:
+#        node_type: 'LoginNode'
+#        head_node_private_ip: '127.0.0.1'
+#        head_node_home_path: '/fake_headnode_home'
+#        shared_dir_head: '/fake_headnode_shared'
   - name: raid_compute
     run_list:
       - recipe[aws-parallelcluster-tests::setup]
@@ -541,7 +542,7 @@ suites:
   - name: shared_storages_compute
     run_list:
       - recipe[aws-parallelcluster-tests::setup]
-      - recipe[aws-parallelcluster-environment::shared_storages]
+      - recipe[aws-parallelcluster-environment::mount_cx_fs]
     verifier:
       controls:
         - shared_storages_compute_and_login
@@ -555,7 +556,7 @@ suites:
   - name: shared_storages_login
     run_list:
       - recipe[aws-parallelcluster-tests::setup]
-      - recipe[aws-parallelcluster-environment::shared_storages]
+      - recipe[aws-parallelcluster-environment::mount_cx_fs]
     verifier:
       controls:
         - shared_storages_compute_and_login

diff --git a/cookbooks/aws-parallelcluster-environment/recipes/config.rb b/cookbooks/aws-parallelcluster-environment/recipes/config.rb
@@ -18,9 +18,9 @@
   action :configure
 end
 include_recipe 'aws-parallelcluster-environment::ephemeral_drives'
-# fs_update generates the shared storages mapping file so must be executed before shared storages recipes
-include_recipe 'aws-parallelcluster-environment::fs_update'
-include_recipe 'aws-parallelcluster-environment::shared_storages'
+# update_fs_mapping generates the shared storages mapping file so must be executed before shared storages recipes
+include_recipe 'aws-parallelcluster-environment::update_fs_mapping'
+include_recipe 'aws-parallelcluster-environment::export_home'
 include_recipe 'aws-parallelcluster-environment::ebs'
 include_recipe 'aws-parallelcluster-environment::raid'
-include_recipe "aws-parallelcluster-environment::fs_mount"
+include_recipe "aws-parallelcluster-environment::mount_cx_fs"
diff --git a/cookbooks/aws-parallelcluster-environment/recipes/config/export_home.rb b/cookbooks/aws-parallelcluster-environment/recipes/config/export_home.rb
@@ -0,0 +1,27 @@
+# frozen_string_literal: true
+
+#
+# Copyright:: 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.
+
+return if on_docker?
+
+case node['cluster']['node_type']
+when 'HeadNode'
+  volume "export /home" do
+    shared_dir "/home"
+    action :export
+  end
+when 'ComputeFleet', 'LoginNode'
+  Chef::Log.info("Export only from the HeadNode")
+else
+  raise "node_type must be HeadNode, ComputeFleet, or LoginNode"
+end
diff --git a/...er-environment/recipes/config/fs_mount.rb → ...environment/recipes/config/mount_cx_fs.rb b/...er-environment/recipes/config/fs_mount.rb → ...environment/recipes/config/mount_cx_fs.rb
@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-# Copyright:: 2013-2022 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+# Copyright:: 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
@@ -23,12 +23,11 @@
 
 # Identify the customer use filesystems and store their data in arrays for the EFS resource
 efs_shared_dir_array.each_with_index do |dir, index|
-  unless node['cluster']['internal_shared_dirs'].include?(dir)
-    cx_shared_dir_array.push(dir)
-    cx_efs_fs_id_array.push(efs_fs_id_array[index])
-    cx_efs_encryption_array.push(efs_encryption_in_transit_array[index])
-    cx_efs_iam_array.push(efs_iam_authorization_array[index])
-  end
+  next if node['cluster']['internal_shared_dirs'].include?(dir)
+  cx_shared_dir_array.push(dir)
+  cx_efs_fs_id_array.push(efs_fs_id_array[index])
+  cx_efs_encryption_array.push(efs_encryption_in_transit_array[index])
+  cx_efs_iam_array.push(efs_iam_authorization_array[index])
 end
 
 # Mount EFS directories with the efs resource

diff --git a/cookbooks/aws-parallelcluster-environment/recipes/config/shared_storages.rb b/cookbooks/aws-parallelcluster-environment/recipes/config/shared_storages.rb
diff --git a/...r-environment/recipes/config/fs_update.rb → ...nment/recipes/config/update_fs_mapping.rb b/...r-environment/recipes/config/fs_update.rb → ...nment/recipes/config/update_fs_mapping.rb
diff --git a/cookbooks/aws-parallelcluster-environment/recipes/init.rb b/cookbooks/aws-parallelcluster-environment/recipes/init.rb
@@ -16,13 +16,11 @@
 cloudwatch "Configure CloudWatch" do
   action :configure
 end
-include_recipe "aws-parallelcluster-environment::fs_update"
+include_recipe "aws-parallelcluster-environment::update_fs_mapping"
 include_recipe "aws-parallelcluster-environment::backup_internal_use_shared_data"
 include_recipe "aws-parallelcluster-environment::mount_internal_use_fs"
 include_recipe "aws-parallelcluster-environment::restore_internal_use_shared_data"
 
-# include_recipe "aws-parallelcluster-environment::mount_shared"
-
 include_recipe "aws-parallelcluster-environment::network_interfaces"
 include_recipe 'aws-parallelcluster-environment::imds'
 

diff --git a/cookbooks/aws-parallelcluster-environment/recipes/init/backup_internal_use_shared_data.rb b/cookbooks/aws-parallelcluster-environment/recipes/init/backup_internal_use_shared_data.rb
@@ -14,24 +14,16 @@
 
 return if on_docker?
 
-case node['cluster']['node_type']
-when 'HeadNode'
-  Chef::Log.info("Internal Shared Storage #{node['cluster']['internal_shared_dirs']}")
+if node['cluster']['node_type'] == 'HeadNode'
   # For each, backup the data to a temp location
   node['cluster']['internal_shared_dirs'].each do |dir|
     bash "Backup #{dir}" do
       user 'root'
       group 'root'
       code <<-EOH
         mkdir -p /tmp#{dir}
-        rsync -a #{dir}/ /tmp#{dir}
+        rsync -a -X #{dir}/ /tmp#{dir}
       EOH
     end
-
   end
-
-when 'ComputeFleet', 'LoginNode'
-  Chef::Log.info("Backup only from the HeadNode")
-else
-  raise "node_type must be HeadNode, ComputeFleet, or LoginNode"
 end
diff --git a/cookbooks/aws-parallelcluster-environment/recipes/init/mount_internal_use_fs.rb b/cookbooks/aws-parallelcluster-environment/recipes/init/mount_internal_use_fs.rb
@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-# Copyright:: 2013-2022 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+# Copyright:: 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
@@ -23,12 +23,11 @@
 
 # Identify the internal use filesystems and store their data in arrays for the EFS resource
 efs_shared_dir_array.each_with_index do |dir, index|
-  if node['cluster']['internal_shared_dirs'].include?(dir)
-    internal_shared_dir_array.push(dir)
-    internal_efs_fs_id_array.push(efs_fs_id_array[index])
-    internal_efs_encryption_array.push(efs_encryption_in_transit_array[index])
-    internal_efs_iam_array.push(efs_iam_authorization_array[index])
-  end
+  next unless node['cluster']['internal_shared_dirs'].include?(dir)
+  internal_shared_dir_array.push(dir)
+  internal_efs_fs_id_array.push(efs_fs_id_array[index])
+  internal_efs_encryption_array.push(efs_encryption_in_transit_array[index])
+  internal_efs_iam_array.push(efs_iam_authorization_array[index])
 end
 
 # Mount EFS directories with the efs resource
@@ -41,7 +40,7 @@
   not_if { internal_shared_dir_array.empty? }
 end
 
-# TODO replace home as NFS with shared /home
+# TODO: replace home as NFS with shared /home
 case node['cluster']['node_type']
 when 'ComputeFleet', 'LoginNode'
   include_recipe 'aws-parallelcluster-environment::mount_home'

diff --git a/cookbooks/aws-parallelcluster-environment/recipes/init/mount_shared.rb b/cookbooks/aws-parallelcluster-environment/recipes/init/mount_shared.rb