Use checkpoint system

Author: hgreebe

cookbooks/aws-parallelcluster-computefleet/recipes/config/config_check_update_systemd_service.rb

@@ -38,6 +38,14 @@
   action :create_if_missing
 end
 
+file node['cluster']['update_checkpoint'] do
+  content ''
+  owner 'root'
+  group 'root'
+  mode '0644'
+  action :create_if_missing
+end
+
 service 'check-update.timer' do
   action [:enable, :start]
 end

cookbooks/aws-parallelcluster-computefleet/templates/check_update/check-update.service.erb

@@ -3,8 +3,18 @@ Description=Check for recent file modifications
 
 [Service]
 Type=oneshot
-ExecStart=/bin/bash -c "echo 'Checking file: <%= node['cluster']['shared_update_path'] %>'; if [ $(( $(date +%%s) - $(stat -c %%Y <%= node['cluster']['shared_update_path'] %>) )) -lt 80 ]; then echo 'File modified recently, running action'; <%= node['cluster']['scripts_dir'] %>/cfn-hup-update-action.sh; else echo 'File not modified recently'; fi"
-
+ExecStart=/bin/bash -c '\
+SHARED_FILE="<%= node['cluster']['shared_update_path'] %>"; \
+LOCAL_CHECKPOINT="<%= node['cluster']['update_checkpoint'] %>"; \
+\
+[ ! -f "$SHARED_FILE" ] && exit 0; \
+\
+CURRENT_UPDATE=$(cat "$SHARED_FILE"); \
+LAST_APPLIED=$([ -f "$LOCAL_CHECKPOINT" ] && cat "$LOCAL_CHECKPOINT" || echo ""); \
+\
+if [ "$CURRENT_UPDATE" != "$LAST_APPLIED" ]; then \
+echo "$CURRENT_UPDATE" > "$LOCAL_CHECKPOINT" && <%= node['cluster']['scripts_dir'] %>/cfn-hup-update-action.sh; \
+fi'
 
 [Install]
 WantedBy=multi-user.target

cookbooks/aws-parallelcluster-environment/templates/cfn_hup_configuration/cfn-hook-update.conf.erb

@@ -3,6 +3,6 @@ triggers=post.update
 <% case node['cluster']['node_type'] -%>
 <% when 'HeadNode', 'LoginNode' -%>
 path=Resources.<%= @launch_template_resource_id %>.Metadata.AWS::CloudFormation::Init
-action=<% if node['cluster']['node_type'] == 'HeadNode' -%>echo "Update triggered at $(date) by <%= node['cluster']['node_type'] %>" >> <%= @update_dir %>; <% end -%>PATH=/usr/local/bin:/bin:/usr/bin:/opt/aws/bin;. /etc/parallelcluster/pcluster_cookbook_environment.sh; $CFN_BOOTSTRAP_VIRTUALENV_PATH/cfn-init -v --stack <%= @stack_id %> --resource <%= @launch_template_resource_id %> --configsets update --region <%= @region %> --url <%= @cloudformation_url %> --role <%= @cfn_init_role %>
+action=<% if node['cluster']['node_type'] == 'HeadNode' -%>echo "Update at $(date)" >> <%= @update_dir %>; <% end -%>PATH=/usr/local/bin:/bin:/usr/bin:/opt/aws/bin;. /etc/parallelcluster/pcluster_cookbook_environment.sh; $CFN_BOOTSTRAP_VIRTUALENV_PATH/cfn-init -v --stack <%= @stack_id %> --resource <%= @launch_template_resource_id %> --configsets update --region <%= @region %> --url <%= @cloudformation_url %> --role <%= @cfn_init_role %>
 <% end %>
 runas=root

cookbooks/aws-parallelcluster-shared/attributes/cluster.rb

@@ -11,6 +11,7 @@
 
 # Shared file used to manage inplace updates
 default['cluster']['shared_update_path'] = "#{node['cluster']['shared_dir']}/check_update"
+default['cluster']['update_checkpoint'] = "#{node['cluster']['scripts_dir']}/update_checkpoint"
 
 # Slurm_plugin_dir is used by slurm cookbook and custom_actions recipe
 default['cluster']['slurm_plugin_dir'] = "#{node['cluster']['etc_dir']}/slurm_plugin"