Merge branch 'develop' into wip/mgiacomo/390/disable-unused-services

Author: gmarciani

CHANGELOG.md

@@ -35,7 +35,6 @@ This file is used to list changes made in each version of the AWS ParallelCluste
 **BUG FIXES**
 - Fix issue making job fail when submitted as active directory user from login nodes. 
   The issue was caused by an incomplete configuration of the integration with the external Active Directory on the head node.
-  This fix comes with a breaking change: now cluster creation/update would fail if the integration with the Active Directory does not work.
 
 3.8.0
 ------

cookbooks/aws-parallelcluster-entrypoints/recipes/finalize.rb

@@ -21,6 +21,7 @@
 end
 
 include_recipe "aws-parallelcluster-platform::finalize"
-include_recipe "aws-parallelcluster-environment::finalize"
 
 include_recipe 'aws-parallelcluster-slurm::finalize' if node['cluster']['scheduler'] == 'slurm'
+
+include_recipe "aws-parallelcluster-environment::finalize"

cookbooks/aws-parallelcluster-environment/recipes/finalize/finalize_directory_service.rb

@@ -22,10 +22,13 @@
   read_only_user = domain_service_read_only_user_name(node['cluster']['directory_service']['domain_read_only_user'])
 
   execute 'Fetch user data from remote directory service' do
-    # The switch-user (sudo -u) is necessary to trigger the fetching of AD data
+    # The switch-user (sudo -u) is necessary to trigger the fetching of AD data.
+    # Failures are ignored because we experimentally verified that a MsAD backend
+    # may take long time to become available.
+    # So, we prefer to execute this step in best effort mode.
+    # Once we will reintroduce the failures, we should consider 30 retries with 10 seconds delay.
     command "sudo -u #{default_user} getent passwd #{read_only_user}"
     user 'root'
-    retries 10 # Retries are just a safe guard in case the node is still fetching data from the AD
-    retry_delay 3
+    ignore_failure true
   end
 end

cookbooks/aws-parallelcluster-environment/spec/unit/recipes/finalize_directory_service_spec.rb

@@ -38,8 +38,7 @@
               is_expected.to run_execute('Fetch user data from remote directory service').with(
                 command: "sudo -u #{cluster_user} getent passwd #{domain_read_only_user}",
                 user: 'root',
-                retries: 10,
-                retry_delay: 3
+                ignore_failure: true
               )
             end
           else

cookbooks/aws-parallelcluster-environment/templates/cfn_bootstrap/cfn-hup-runner.sh.erb

@@ -10,5 +10,5 @@ set -ex
 
 while true; do
   <%= @cfn_bootstrap_virtualenv_path %>/bin/cfn-hup --no-daemon --verbose
-  sleep 120
+  sleep 60
 done

cookbooks/aws-parallelcluster-platform/resources/nvidia_driver/partial/_nvidia_driver_common.rb

@@ -23,6 +23,10 @@
   return unless nvidia_driver_enabled?
   return if on_docker?
 
+  # Share nvidia driver version with InSpec tests
+  node.default['cluster']['nvidia']['driver_version'] = _nvidia_driver_version
+  node_attributes "Save Nvidia driver version for Inspec tests"
+
   remote_file tmp_nvidia_run do
     source nvidia_driver_url
     mode '0755'

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

@@ -212,6 +212,8 @@ def self.configure(chef_run)
           'centos7' => 'el7',
           'rhel8' => 'el8',
           'rocky8' => 'el8',
+          'rhel9' => 'el9',
+          'rocky9' => 'el9',
           'ubuntu20.04' => 'Ubuntu20_04',
           'ubuntu22.04' => 'Ubuntu22_04',
         }

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

@@ -187,6 +187,10 @@ def self.setup(chef_run, nvidia_driver_version: nil)
         end
         cached(:node) { chef_run.node }
 
+        it 'dumps nodes attribues' do
+          is_expected.to write_node_attributes('Save Nvidia driver version for Inspec tests')
+        end
+
         it 'sets up nvidia_driver' do
           is_expected.to setup_nvidia_driver('setup')
         end

util/upload-cookbook.sh

@@ -24,6 +24,7 @@ _help() {
   --profile <aws-profile>       AWS profile name to use for the upload
                                 (optional, default is AWS_PROFILE env variable or "default")
   --region <aws-region>         Region to use for AWSCli commands (optional, default is "us-east-1")
+  --scope <string>              Disambiguation string used in the S3 path to avoid collisions (default is empty)
   -h, --help                    Print this help message
 EOF
 }
@@ -40,6 +41,8 @@ main() {
             --profile=*)        _profile="${1#*=}";;
             --region)           _region="$2"; shift;;
             --region=*)         _region="${1#*=}";;
+            --scope)            _scope="$2"; shift;;
+            --scope=*)          _scope="${1#*=}";;
             -h|--help|help)     _help; exit 0;;
             *)                  _help; echo "[error] Unrecognized option '$1'"; exit 1;;
         esac
@@ -66,6 +69,10 @@ main() {
         _info "--region parameter not specified, using 'us-east-1'"
         _region="us-east-1"
     fi
+    if [ -z "${_scope}" ]; then
+        _info "--scope parameter not specified, no scope will be used"
+        _scope=""
+    fi
 
     # check bucket or create it
     aws ${_profile} s3api head-bucket --bucket "${_bucket}" --region "${_region}"
@@ -94,6 +101,9 @@ main() {
 
     # upload packages
     _key_path="parallelcluster/${_version}/cookbooks"
+    if [ -n "${_scope}" ]; then
+        _key_path="${_key_path}/${_scope}"
+    fi
     aws ${_profile} --region "${_region}" s3 cp aws-parallelcluster-cookbook-${_version}.tgz s3://${_bucket}/${_key_path}/aws-parallelcluster-cookbook-${_version}.tgz || _error_exit 'Failed to push cookbook to S3'
     aws ${_profile} --region "${_region}" s3 cp aws-parallelcluster-cookbook-${_version}.md5 s3://${_bucket}/${_key_path}/aws-parallelcluster-cookbook-${_version}.md5 || _error_exit 'Failed to push cookbook md5 to S3'
     aws ${_profile} --region "${_region}" s3api head-object --bucket ${_bucket} --key ${_key_path}/aws-parallelcluster-cookbook-${_version}.tgz --output text --query LastModified > aws-parallelcluster-cookbook-${_version}.tgz.date || _error_exit 'Failed to fetch LastModified date'