Add unit and kitchen tests

CHANGELOG.md

@@ -10,6 +10,7 @@ This file is used to list changes made in each version of the AWS ParallelCluste
 - Allow custom actions on login nodes.
 - Allow DCV connection on login nodes.
 - Add new attribute `efs_access_point_ids` to specify optional EFS access points for the mounts
+- Install enroot and pyxis in official pcluster AMIs
 
 **BUG FIXES**
 - Fix EFA kmod installation with RHEL 8.10 or newer.

cookbooks/aws-parallelcluster-platform/resources/enroot/partial/_enroot_common.rb

@@ -16,11 +16,12 @@
 default_action :setup
 
 action :setup do
+  return if on_docker?
   action_install_package
 end
 
 action :configure do
-  enroot_installed = ::File.exist?('/usr/bin/enroot')
+  return if on_docker?
   return unless enroot_installed
 
   bash "Configure enroot" do
@@ -36,14 +37,14 @@
       ENROOT_CACHE_PATH=${SHARED_DIR}/enroot envsubst < /tmp/enroot.template.conf > /tmp/enroot.conf
       mv /tmp/enroot.conf /etc/enroot/enroot.conf
       chmod 0644 /etc/enroot/enroot.conf
-  
+
       mkdir -p /tmp/enroot
       chmod 1777 /tmp/enroot
       mkdir -p /tmp/enroot/data
       chmod 1777 /tmp/enroot/data
-  
+
       chmod 1777 ${SHARED_DIR}/enroot
-  
+
       mkdir -p ${SHARED_DIR}/pyxis/
       chown ${NONROOT_USER} ${SHARED_DIR}/pyxis/
       sed -i '${s/$/ runtime_path=${SHARED_DIR}\\/pyxis/}' /opt/slurm/etc/plugstack.conf.d/pyxis.conf
@@ -59,3 +60,7 @@
 def package_version
   node['cluster']['enroot']['version']
 end
+
+def enroot_installed
+  ::File.exist?('/usr/bin/enroot')
+end

cookbooks/aws-parallelcluster-platform/resources/enroot/partial/_enroot_debian.rb

@@ -24,7 +24,7 @@
       curl -fSsL -O #{enroot_url}
       curl -fSsL -O #{enroot_caps_url}
       apt install -y ./*.deb
-  
+
       ln -s /usr/share/enroot/hooks.d/50-slurm-pmi.sh /etc/enroot/hooks.d/
       ln -s /usr/share/enroot/hooks.d/50-slurm-pytorch.sh /etc/enroot/hooks.d/
       mkdir -p /etc/sysconfig

cookbooks/aws-parallelcluster-platform/spec/unit/resources/enroot_spec.rb

@@ -0,0 +1,171 @@
+require 'spec_helper'
+
+package_version = '3.4.1'
+class ConvergeEnroot
+  def self.setup(chef_run)
+    chef_run.converge_dsl('aws-parallelcluster-platform') do
+      enroot 'setup' do
+        action :setup
+      end
+    end
+  end
+
+  def self.configure(chef_run)
+    chef_run.converge_dsl('aws-parallelcluster-platform') do
+      enroot 'configure' do
+        action :configure
+      end
+    end
+  end
+end
+
+describe 'enroot:package_version' do
+  for_all_oses do |platform, version|
+    context "on #{platform}#{version}" do
+      cached(:chef_run) do
+        allow_any_instance_of(Object).to receive(:nvidia_enabled?).and_return(false)
+        runner = runner(platform: platform, version: version, step_into: ['enroot'])
+        ConvergeEnroot.setup(runner)
+      end
+      cached(:resource) do
+        chef_run.find_resource('enroot', 'setup')
+      end
+
+      it 'returns the expected enroot version' do
+        expected_enroot_version = "3.4.1"
+        expect(resource.package_version).to eq(expected_enroot_version)
+      end
+    end
+  end
+end
+
+describe 'enroot:arch_suffix' do
+  for_all_oses do |platform, version|
+    context "on #{platform}#{version} - arm" do
+      cached(:chef_run) do
+        allow_any_instance_of(Object).to receive(:nvidia_enabled?).and_return(false)
+        runner = runner(platform: platform, version: version, step_into: ['enroot'])
+        ConvergeEnroot.setup(runner)
+      end
+      cached(:resource) do
+        chef_run.find_resource('enroot', 'setup')
+      end
+
+      context 'on arm instance' do
+        cached(:expected_arch) do
+          case platform
+          when 'amazon', 'redhat', 'rocky'
+            'aarch64'
+          else
+            'arm64'
+          end
+        end
+
+        it 'returns arch value for arm architecture' do
+          allow_any_instance_of(Object).to receive(:arm_instance?).and_return(true)
+          expect(resource.arch_suffix).to eq(expected_arch)
+        end
+      end
+
+      context 'not on arm instance' do
+        cached(:expected_arch) do
+          platform == 'ubuntu' ? 'amd64' : 'x86_64'
+        end
+
+        it 'returns arch value for arm architecture' do
+          allow_any_instance_of(Object).to receive(:arm_instance?).and_return(false)
+          expect(resource.arch_suffix).to eq(expected_arch)
+        end
+      end
+    end
+  end
+end
+
+describe 'enroot:setup' do
+  for_all_oses do |platform, version|
+    context "on #{platform}#{version}" do
+      let(:chef_run) do
+        runner(platform: platform, version: version, step_into: ['enroot']) do |node|
+          node.override['cluster']['enroot']['version'] = package_version
+        end
+      end
+
+      context 'when nvidia is enabled' do
+        before do
+          stubs_for_provider('enroot') do |resource|
+            allow(resource).to receive(:nvidia_enabled?).and_return(true)
+          end
+        end
+
+        context 'and enroot is installed' do
+          before do
+            ConvergeEnroot.setup(chef_run)
+          end
+
+          it 'installs Enroot' do
+            is_expected.to run_bash('Install enroot')
+          end
+        end
+      end
+
+      context 'when nvidia is not enabled' do
+        before do
+          stubs_for_provider('enroot') do |resource|
+            allow(resource).to receive(:nvidia_enabled?).and_return(false)
+          end
+        end
+
+        context 'and enroot is installed' do
+          before do
+            ConvergeEnroot.setup(chef_run)
+          end
+
+          it 'does not install Enroot' do
+            is_expected.not_to run_bash('Install enroot')
+          end
+        end
+      end
+    end
+  end
+end
+
+describe 'enroot:configure' do
+  for_all_oses do |platform, version|
+    context "on #{platform}#{version}" do
+      let(:chef_run) do
+        runner(platform: platform, version: version, step_into: ['enroot'])
+      end
+
+      context 'when enroot is installed' do
+        before do
+          stubs_for_provider('enroot') do |resource|
+            allow(resource).to receive(:enroot_installed).and_return(true)
+          end
+          ConvergeEnroot.configure(chef_run)
+        end
+        it 'run configure enroot script' do
+          is_expected.to run_bash('Configure enroot')
+            .with(retries: 3)
+            .with(retry_delay: 5)
+            .with(user: 'root')
+        end
+      end
+
+      context 'when enroot is not installed' do
+        before do
+          stubs_for_provider('enroot') do |resource|
+            allow(resource).to receive(:enroot_installed).and_return(false)
+          end
+          ConvergeEnroot.configure(chef_run)
+        end
+
+        it 'does not run configure enroot script' do
+          is_expected.not_to run_bash('Configure enroot')
+            .with(retries: 3)
+            .with(retry_delay: 5)
+            .with(user: 'root')
+        end
+      end
+    end
+  end
+end

cookbooks/aws-parallelcluster-platform/test/controls/enroot_spec.rb

@@ -0,0 +1,47 @@
+# 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.
+
+control 'tag:install_expected_version_of_enroot_installed' do
+  only_if { !os_properties.on_docker? && ['yes', true].include?(node['cluster']['nvidia']['enabled']) }
+
+  expected_enroot_version = node['cluster']['enroot']['version']
+
+  describe "gdrcopy version is expected to be #{expected_enroot_version}" do
+    subject { command('enroot version').stdout.strip() }
+    it { should eq expected_enroot_version }
+  end
+end
+
+control 'tag:config_enroot_enabled_on_graphic_instances' do
+  only_if { !os_properties.on_docker? && ['yes', true].include?(node['cluster']['nvidia']['enabled']) }
+
+  describe directory('/usr/bin/enroot') do
+    it { should exist }
+    its('owner') { should eq 'root' }
+    its('group') { should eq 'root' }
+  end
+
+  describe file("/opt/parallelcluster/shared/enroot") do
+    it { should exist }
+    its('mode') { should cmp '1777' }
+    its('owner') { should eq 'root' }
+    its('group') { should eq 'root' }
+  end unless os_properties.redhat_on_docker?
+end
+
+control 'tag:config_enroot_disabled_on_non_graphic_instances' do
+  only_if { !os_properties.on_docker? && !['yes', true].include?(node['cluster']['nvidia']['enabled']) }
+
+  describe 'enroot service should be disabled' do
+    subject { command("enroot version") }
+    its('exit_status') { should eq 127 }
+  end
+end

cookbooks/aws-parallelcluster-slurm/test/controls/pyxis_spec.rb

@@ -0,0 +1,21 @@
+# 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.
+
+control 'tag:install_pyxis_installed' do
+  title 'Checks Pyxis has been installed'
+
+  describe file("/opt/slurm/etc/plugstack.conf.d/pyxis.conf") do
+    it { should exist }
+    its('mode') { should cmp '0755' }
+    its('owner') { should eq 'root' }
+    its('group') { should eq 'root' }
+  end unless os_properties.redhat_on_docker?
+end