Remove prolog/epilog used as a workaround for cluster without Internet connection.

The Slurm bug has been fixed with Slurm 22.05. The workaround is no longer necessary

Signed-off-by: Hanwen <[email protected]>

Author: hanwen-cluster

attributes/default.rb

@@ -552,7 +552,6 @@
 default['cluster']['raid_vol_ids'] = ''
 default['cluster']['dns_domain'] = nil
 default['cluster']['use_private_hostname'] = 'false'
-default['cluster']['add_node_hostnames_in_hosts_file'] = node['cluster']['use_private_hostname']
 default['cluster']['skip_install_recipes'] = 'yes'
 default['cluster']['enable_nss_slurm'] = node['cluster']['directory_service']['enabled']
 default['cluster']['realmemory_to_ec2memory_ratio'] = 0.95

cookbooks/aws-parallelcluster-config/files/default/cloudwatch_agent/cloudwatch_log_files.json

@@ -540,28 +540,6 @@
       ],
       "feature_conditions": []
     },
-    {
-      "timestamp_format_key": "default",
-      "file_path": "/var/log/parallelcluster/slurm_prolog_epilog.log",
-      "log_stream_name": "slurm_prolog_epilog",
-      "schedulers": [
-        "slurm"
-      ],
-      "platforms": [
-        "centos",
-        "ubuntu",
-        "amazon"
-      ],
-      "node_roles": [
-        "ComputeFleet"
-      ],
-      "feature_conditions": [
-        {
-          "dna_key": "use_private_hostname",
-          "satisfying_values": ["true"]
-        }
-      ]
-    },
     {
       "timestamp_format_key": "default",
       "file_path": "/var/log/parallelcluster/clusterstatusmgtd",

cookbooks/aws-parallelcluster-slurm/files/default/head_node_slurm/epilog

@@ -192,34 +192,6 @@
   action :create
 end
 
-if node['cluster']['add_node_hostnames_in_hosts_file'] == "true"
-  directory "#{node['cluster']['slurm']['install_dir']}/etc/pcluster/prolog.d" do
-    user 'root'
-    group 'root'
-    mode '0755'
-  end
-
-  cookbook_file "#{node['cluster']['slurm']['install_dir']}/etc/pcluster/prolog.d/01-pcluster-prolog" do
-    source 'head_node_slurm/prolog'
-    owner node['cluster']['slurm']['user']
-    group node['cluster']['slurm']['group']
-    mode '0744'
-  end
-
-  directory "#{node['cluster']['slurm']['install_dir']}/etc/pcluster/epilog.d" do
-    user 'root'
-    group 'root'
-    mode '0755'
-  end
-
-  cookbook_file "#{node['cluster']['slurm']['install_dir']}/etc/pcluster/epilog.d/01-pcluster-epilog" do
-    source 'head_node_slurm/epilog'
-    owner node['cluster']['slurm']['user']
-    group node['cluster']['slurm']['group']
-    mode '0744'
-  end
-end
-
 service "slurmctld" do
   supports restart: false
   action %i(enable start)

cookbooks/aws-parallelcluster-slurm/files/default/head_node_slurm/prolog

@@ -40,19 +40,6 @@ SuspendTimeout=120
 PrivateData=cloud
 ResumeRate=0
 SuspendRate=0
-<% if node["cluster"]["add_node_hostnames_in_hosts_file"] == 'true' -%>
-#
-# PROLOG AND EPILOG
-# prolog is executed to add nodes info to /etc/hosts on compute nodes when each job is allocated
-# epilog is executed to clean contents written by prolog
-# PrologFlags specifies the prolog is executed at job allocation and prologs and epilogs are of different jobs are executed serially
-# SchedulerParameters allows jobs to be requeued to other nodes if prolog error exits.
-# Note the error exit of prolog drains a node, because the error of prolog is considered as a node error.
-Epilog=<%= node['cluster']['slurm']['install_dir'] %>/etc/pcluster/epilog.d/*
-Prolog=<%= node['cluster']['slurm']['install_dir'] %>/etc/pcluster/prolog.d/*
-PrologFlags=alloc,serial
-SchedulerParameters=nohold_on_prolog_fail
-<% end -%>
 #
 # TIMERS
 SlurmctldTimeout=300