Modify the check_login_stopped.sh script to directly use TG name to q…

…uery (#2524)

Reduce the check steps when checking if LoginNodes stopped. 
So that we can delete the checking NLB loop to reduce running time.

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

@@ -15,6 +15,9 @@
 #
 # Retrieve compute nodename from file
 #
+
+require 'digest'
+
 def slurm_nodename
   slurm_nodename_file = "#{node['cluster']['slurm_plugin_dir']}/slurm_nodename"
 
@@ -137,3 +140,11 @@ def get_primary_ip
 
   primary_ip
 end
+
+def get_target_group_name(cluster_name, pool_name)
+  partial_cluster_name = cluster_name[0..6]
+  partial_pool_name = pool_name[0..6]
+  combined_name = cluster_name + pool_name
+  hash_value = Digest::SHA256.hexdigest(combined_name)[0..15]
+  "#{partial_cluster_name}-#{partial_pool_name}-#{hash_value}"
+end

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

@@ -11,14 +11,19 @@
 # OR CONDITIONS OF ANY KIND, express or implied. See the License for the specific language governing permissions and
 # limitations under the License.
 # rubocop:disable Style/SingleArgumentDig
+
 template "#{node['cluster']['scripts_dir']}/slurm/check_login_nodes_stopped.sh" do
   source 'slurm/head_node/check_login_nodes_stopped.sh.erb'
   owner 'root'
   group 'root'
   mode '0700'
   variables(
-    cluster_name: node['cluster']['cluster_name'] || node['cluster']['stack_name'],
-    login_nodes_pool_name: lazy { node['cluster']['config'].dig(:LoginNodes, :Pools, 0, :Name) },
+    target_group_name: lazy do
+      get_target_group_name(
+        node['cluster']['cluster_name'] || node['cluster']['stack_name'],
+        node['cluster']['config'].dig(:LoginNodes, :Pools, 0, :Name)
+      )
+    end,
     region: node['cluster']['region']
   )
   only_if do

cookbooks/aws-parallelcluster-slurm/spec/unit/libraries/get_target_group_name_spec.rb

@@ -0,0 +1,22 @@
+require_relative '../../../libraries/helpers'
+
+describe 'get_target_group_name' do
+  shared_examples 'a valid target group name generator' do |cluster_name, pool_name, expected_result|
+    it 'generates a correctly formatted target group name' do
+      target_group_name = get_target_group_name(cluster_name, pool_name)
+      expect(target_group_name).to eq(expected_result)
+    end
+  end
+
+  context 'when cluster and pool names are regular strings' do
+    include_examples 'a valid target group name generator', 'test-cluster', 'test-pool', 'test-cl-test-po-18c74b16dfbc78ac'
+  end
+
+  context 'when cluster and pool names are longer strings' do
+    include_examples 'a valid target group name generator', 'abcdefghijklmnopqrstuvwxyz', 'zyxwvutsrqponmlkjihgfedcba', 'abcdefg-zyxwvut-20f1fcdf919164c7'
+  end
+
+  context 'when cluster and pool names are single characters' do
+    include_examples 'a valid target group name generator', 'a', 'b', 'a-b-fb8e20fc2e4c3f24'
+  end
+end

cookbooks/aws-parallelcluster-slurm/templates/default/slurm/head_node/check_login_nodes_stopped.sh.erb

@@ -1,50 +1,19 @@
 #!/bin/bash
 # This script checks whether there are running login nodes in a specified AWS ParallelCluster stack and login nodes pool.
-# It first retrieves the ARN of the Load Balancer associated with the specified stack and login nodes pool.
-# If a Load Balancer is found, it then retrieves the ARN of the Target Group associated with the Load Balancer.
-# Lastly, it checks the health of the targets in the Target Group to determine the number of healthy and unhealthy login nodes.
+# It first retrieves the ARN of the Target Group associated with the LoginNodes Network Load Balancer.
+# Then, it checks the health of the targets in the Target Group to determine the number of healthy and unhealthy login nodes.
 # If there are any healthy or unhealthy nodes found, it concludes that there are running login nodes.
 #
 # Usage: ./check_if_has_running_login_nodes.sh
 
 set -e
 
-CLUSTER_NAME="<%= @cluster_name %>"
-LOGIN_NODES_POOL_NAME="<%= @login_nodes_pool_name %>"
+TARGET_GROUP_NAME="<%= @target_group_name %>"
 REGION="<%= @region %>"
 
-# List all Load Balancers
-load_balancers=$(aws elbv2 describe-load-balancers --region ${REGION})
-
-# Iterate over Load Balancers to find the one with matching tags
-load_balancer_arn=''
-for arn in $(echo "${load_balancers}" | jq -r '.LoadBalancers[].LoadBalancerArn'); do
-    # Get tags for the current Load Balancer
-    tags=$(aws elbv2 describe-tags --resource-arns "${arn}" --region ${REGION})
-
-    # Check if the tags match the desired stack name and login nodes pool name
-    cluster_name_match=$(echo "${tags}" | jq -r --arg key "parallelcluster:cluster-name" --arg value "${CLUSTER_NAME}" '.TagDescriptions[] | select(.Tags[]? | (.Key == $key and .Value == $value))')
-    login_nodes_pool_name_match=$(echo "${tags}" | jq -r --arg key "parallelcluster:login-nodes-pool" --arg value "${LOGIN_NODES_POOL_NAME}" '.TagDescriptions[] | select(.Tags[]? | (.Key == $key and .Value == $value))')
-
-    # If both tags are found, store the ARN and break the loop
-    # For now, there's only one pool of login nodes per cluster.
-    if [[ -n "${cluster_name_match}" && -n "${login_nodes_pool_name_match}" ]]; then
-        load_balancer_arn="${arn}"
-        break
-    fi
-done
-
-# Output result
-if [[ -n "${load_balancer_arn}" ]]; then
-    echo "Load Balancer ARN found: ${load_balancer_arn}"
-else
-    echo "No Load Balancer found for the cluster ${CLUSTER_NAME} and login nodes pool ${LOGIN_NODES_POOL_NAME}."
-    exit 1
-fi
-
-# Get Target Group ARN associated with the Load Balancer
+# Get Target Group ARN
 target_group_arn=$(aws elbv2 describe-target-groups \
-    --load-balancer-arn $load_balancer_arn \
+    --names ${TARGET_GROUP_NAME} \
     --query "TargetGroups[0].TargetGroupArn" \
     --output text \
     --region ${REGION})