Merge branch 'develop' into wip/munge-key-rotation

CHANGELOG.md

@@ -14,6 +14,13 @@ This file is used to list changes made in each version of the AWS ParallelCluste
 **BUG FIXES**
 - Fix inconsistent scaling configuration after cluster update rollback when modifying the list of instance types declared in the Compute Resources.
 
+3.7.1
+------
+**CHANGES**
+- Upgrade Slurm to 23.02.5 (from 23.02.4).
+  - Upgrade Pmix to 4.2.6 (from 3.2.3).
+  - Upgrade libjwt to 1.15.3 (from 1.12.0).
+
 3.7.0
 ------
 

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

@@ -65,6 +65,15 @@ def enable_munge_service
   end
 end
 
+def restart_munge_service
+  service "munge" do
+    supports restart: true
+    action :restart
+    retries 5
+    retry_delay 10
+  end
+end
+
 def setup_munge_head_node
   # Generate munge key or get it's value from secrets manager
   munge_key_manager 'manage_munge_key' do
@@ -77,6 +86,17 @@ def setup_munge_head_node
   share_munge_head_node
 end
 
+def update_munge_head_node
+  munge_key_manager 'update_munge_key' do
+    munge_key_secret_arn lazy { node['cluster']['config'].dig(:DevSettings, :SlurmSettings, :MungeKeySecretArn) }
+    action :update_munge_key
+    only_if { ::File.exist?(node['cluster']['previous_cluster_config_path']) && is_custom_munge_key_updated? }
+  end
+
+  restart_munge_service
+  share_munge_head_node
+end
+
 def share_munge_head_node
   # Share munge key
   bash 'share_munge_key' do

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

@@ -53,11 +53,18 @@ def is_compute_node_bootstrap_timeout_updated?(previous_config, config)
   evaluate_compute_bootstrap_timeout(previous_config) != evaluate_compute_bootstrap_timeout(config)
 end
 
-def is_slurm_database_updated?
+def config_parameter_changed?(param)
+  # Compares previous cluster config with the current one for changes in a parameter
+  # Parameters:
+  # - `param`: An array representing the sequence of nested keys to the parameter to be checked
   require 'yaml'
   config = YAML.safe_load(File.read(node['cluster']['cluster_config_path']))
   previous_config = YAML.safe_load(File.read(node['cluster']['previous_cluster_config_path']))
-  config["Scheduling"]["SlurmSettings"]["Database"] != previous_config["Scheduling"]["SlurmSettings"]["Database"]
+  config.dig(*param) != previous_config.dig(*param)
+end
+
+def is_slurm_database_updated?
+  config_parameter_changed?(%w(Scheduling SlurmSettings Database))
 end
 
 def raise_command_error(command, cmd)
@@ -71,3 +78,8 @@ def execute_command(command, user = "root", timeout = 300, raise_on_error = true
   raise_command_error(command, cmd) if raise_on_error && cmd.error?
   cmd.stdout.strip
 end
+
+# Verify if MungeKeySecretArn in SlurmSettings section of cluster configuration has been updated
+def is_custom_munge_key_updated?
+  config_parameter_changed?(%w(DevSettings SlurmSettings MungeKeySecretArn))
+end

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

@@ -217,6 +217,8 @@ def update_nodes_in_queue(strategy, queues)
   only_if { ::File.exist?(node['cluster']['previous_cluster_config_path']) && is_custom_munge_key_updated? }
 end
 
+update_munge_head_node
+
 # The previous execute "generate_pcluster_slurm_configs" block resource may have overridden the slurmdbd password in
 # slurm_parallelcluster_slurmdbd.conf with a default value, so if it has run and Slurm accounting
 # is enabled we must pull the database password from Secrets Manager once again.

cookbooks/aws-parallelcluster-slurm/resources/munge_key_manager.rb

@@ -23,50 +23,67 @@
 
 default_action :setup_munge_key
 
-action :setup_munge_key do
-  if new_resource.munge_key_secret_arn
-    # This block will fetch the munge key from Secrets Manager
-    bash 'fetch_and_decode_munge_key' do
-      user 'root'
-      group 'root'
-      cwd '/tmp'
-      code <<-FETCH_AND_DECODE
-        set -e
-        # Get encoded munge key from secrets manager
-        encoded_key=$(aws secretsmanager get-secret-value --secret-id #{new_resource.munge_key_secret_arn} --query 'SecretString' --output text --region #{node['cluster']['region']})
-        # If encoded_key doesn't have a value, error and exit
-        if [ -z "$encoded_key" ]; then
-          echo "Error fetching munge key from Secrets Manager or the key is empty"
-          exit 1
-        fi
+def fetch_and_decode_munge_key
+  declare_resource(:bash, 'fetch_and_decode_munge_key') do
+    user 'root'
+    group 'root'
+    cwd '/tmp'
+    code <<-FETCH_AND_DECODE
+      set -e
+      # Get encoded munge key from secrets manager
+      encoded_key=$(aws secretsmanager get-secret-value --secret-id #{new_resource.munge_key_secret_arn} --query 'SecretString' --output text --region #{node['cluster']['region']})
+      # If encoded_key doesn't have a value, error and exit
+      if [ -z "$encoded_key" ]; then
+        echo "Error fetching munge key from Secrets Manager or the key is empty"
+        exit 1
+      fi
 
-        # Decode munge key and write to /etc/munge/munge.key
-        decoded_key=$(echo $encoded_key | base64 -d)
-        if [ $? -ne 0 ]; then
-          echo "Error decoding the munge key with base64"
-          exit 1
-        fi
+      # Decode munge key and write to /etc/munge/munge.key
+      decoded_key=$(echo $encoded_key | base64 -d)
+      if [ $? -ne 0 ]; then
+        echo "Error decoding the munge key with base64"
+        exit 1
+      fi
 
-        echo "$decoded_key" > /etc/munge/munge.key
+      echo "$decoded_key" > /etc/munge/munge.key
 
-        # Set ownership on the key
-        chown #{node['cluster']['munge']['user']}:#{node['cluster']['munge']['group']} /etc/munge/munge.key
-        # Enforce correct permission on the key
-        chmod 0600 /etc/munge/munge.key
-      FETCH_AND_DECODE
-    end
-  else
-    # This block will generate a munge key if it doesn't exist
-    bash 'generate_munge_key' do
-      not_if { ::File.exist?('/etc/munge/munge.key') }
-      user node['cluster']['munge']['user']
-      group node['cluster']['munge']['group']
-      cwd '/tmp'
-      code <<-GENERATE_KEY
+      # Set ownership on the key
+      chown #{node['cluster']['munge']['user']}:#{node['cluster']['munge']['group']} /etc/munge/munge.key
+      # Enforce correct permission on the key
+      chmod 0600 /etc/munge/munge.key
+    FETCH_AND_DECODE
+  end
+end
+
+def generate_munge_key
+  declare_resource(:bash, 'generate_munge_key') do
+    user node['cluster']['munge']['user']
+    group node['cluster']['munge']['group']
+    cwd '/tmp'
+    code <<-GENERATE_KEY
         set -e
         /usr/sbin/mungekey --verbose
         chmod 0600 /etc/munge/munge.key
-      GENERATE_KEY
-    end
+    GENERATE_KEY
+  end
+end
+
+action :setup_munge_key do
+  if new_resource.munge_key_secret_arn
+    # This block will fetch the munge key from Secrets Manager
+    fetch_and_decode_munge_key
+  else
+    # This block will randomly generate a munge key
+    generate_munge_key
+  end
+end
+
+action :update_munge_key do
+  if new_resource.munge_key_secret_arn
+    # This block will fetch the munge key from Secrets Manager and replace the previous munge key
+    fetch_and_decode_munge_key
+  else
+    # This block will randomly generate a munge key and replace the previous munge key
+    generate_munge_key
   end
 end