Support Slurm 23.11 (#54)

* Support Slurm 23.11 * Fix user shell for Debian * Manage cli_filter.lua like other configless managed configs * Add dirs parameter to slurm::job_container defind type * Add shared parameter to slurm::job_container * Make cgroup_signal_child_processes optional * Add reconfig_ignore_errors parameter * Support newer systemd module * The scrun.lua is now part of configless deployment
treydock · Mar 25, 2024 · e9f9645 · e9f9645
1 parent 8092040
commit e9f9645
Show file tree

Hide file tree

Showing 25 changed files with 173 additions and 117 deletions.
diff --git a/README.md b/README.md
@@ -27,14 +27,15 @@ Manage SLURM.
 
 ### Supported Versions of SLURM
 
-This module is designed to work with SLURM 21.08.x, 22.05.x and 23.02.x.
+This module is designed to work with SLURM 22.05.x, 23.02.x and 23.11.x.
 
 | SLURM Version     | SLURM Puppet module versions |
 | ----------------- | -----------------------------|
 | 20.02.x           | 0.x                          |
 | 20.11.x           | 1.x                          |
 | 21.08.x & 22.05.x | 2.x                          |
 | 23.02.x           | 3.x                          |
+| 23.11.x           | 4.x                          |
 
 ## Usage
 

diff --git a/data/os/Debian.yaml b/data/os/Debian.yaml
@@ -25,3 +25,4 @@ slurm::source_dependencies:
   - libyaml-dev
 slurm::env_dir: /etc/default
 slurm::slurmrestd_user_group: nogroup
+slurm::slurm_user_shell: /usr/sbin/nologin
diff --git a/data/os/Ubuntu.yaml b/data/os/Ubuntu.yaml
@@ -0,0 +1 @@
+slurm::slurm_user_shell: /sbin/nologin
diff --git a/hiera.yaml b/hiera.yaml
@@ -8,6 +8,8 @@ defaults:  # Used for any hierarchy level that omits these keys.
 hierarchy:
   - name: 'os name major release'
     path: "os/%{facts.os.name}/%{facts.os.release.major}.yaml"
+  - name: 'os name'
+    path: "os/%{facts.os.name}.yaml"
   - name: 'os family major release'
     path: "os/%{facts.os.family}/%{facts.os.release.major}.yaml"
   - name: 'os family'

diff --git a/manifests/common/config.pp b/manifests/common/config.pp
@@ -34,9 +34,7 @@
     $slurm::nodesets.each |$name, $_nodeset| {
       slurm::nodeset { $name: * => $_nodeset }
     }
-  }
 
-  if $slurm::manage_slurm_conf and ! $slurm::configless {
     concat { 'slurm-topology.conf':
       ensure => 'present',
       path   => $slurm::topology_conf_path,
@@ -139,35 +137,35 @@
       source  => $slurm::oci_conf_source,
       notify  => $slurm::service_notify,
     }
-  }
 
-  if ($slurm::client or $slurm::slurmctld) and ($slurm::cli_filter_lua_source or $slurm::cli_filter_lua_content) {
-    file { "${slurm::conf_dir}/cli_filter.lua":
-      ensure  => 'file',
-      owner   => 'root',
-      group   => 'root',
-      mode    => '0644',
-      source  => $slurm::cli_filter_lua_source,
-      content => $slurm::cli_filter_lua_content,
-    }
+    if $slurm::cli_filter_lua_source or $slurm::cli_filter_lua_content {
+      file { "${slurm::conf_dir}/cli_filter.lua":
+        ensure  => 'file',
+        owner   => 'root',
+        group   => 'root',
+        mode    => '0644',
+        source  => $slurm::cli_filter_lua_source,
+        content => $slurm::cli_filter_lua_content,
+      }
 
-    if $slurm::slurmctld and $slurm::enable_configless {
-      File["${slurm::conf_dir}/cli_filter.lua"] ~> Exec['scontrol reconfig']
+      if $slurm::slurmctld and $slurm::enable_configless {
+        File["${slurm::conf_dir}/cli_filter.lua"] ~> Exec['scontrol reconfig']
+      }
     }
-  }
 
-  if ($slurm::client or $slurm::slurmctld) and ($slurm::scrun_lua_source or $slurm::scrun_lua_content) {
-    file { "${slurm::conf_dir}/scrun.lua":
-      ensure  => 'file',
-      owner   => 'root',
-      group   => 'root',
-      mode    => '0644',
-      source  => $slurm::scrun_lua_source,
-      content => $slurm::scrun_lua_content,
-    }
+    if $slurm::scrun_lua_source or $slurm::scrun_lua_content {
+      file { "${slurm::conf_dir}/scrun.lua":
+        ensure  => 'file',
+        owner   => 'root',
+        group   => 'root',
+        mode    => '0644',
+        source  => $slurm::scrun_lua_source,
+        content => $slurm::scrun_lua_content,
+      }
 
-    if $slurm::slurmctld and $slurm::enable_configless {
-      File["${slurm::conf_dir}/scrun.lua"] ~> Exec['scontrol reconfig']
+      if $slurm::slurmctld and $slurm::enable_configless {
+        File["${slurm::conf_dir}/scrun.lua"] ~> Exec['scontrol reconfig']
+      }
     }
   }
 

diff --git a/manifests/init.pp b/manifests/init.pp
@@ -37,6 +37,7 @@
 # @param reload_services
 # @param restart_services
 # @param slurmctld_conn_validator_timeout
+# @param reconfig_ignore_errors
 # @param manage_slurm_user
 # @param slurm_user_group
 # @param slurm_group_gid
@@ -145,7 +146,6 @@
 # @param slurmrestd_restart_on_failure
 # @param cgroup_conf_template
 # @param cgroup_conf_source
-# @param cgroup_automount
 # @param cgroup_mountpoint
 # @param cgroup_plugin
 # @param cgroup_allowed_ram_space
@@ -158,6 +158,7 @@
 # @param cgroup_max_swap_percent
 # @param cgroup_memory_swappiness
 # @param cgroup_min_ram_space
+# @param cgroup_signal_child_processes
 # @param oci_conf_template
 # @param oci_conf_source
 # @param oci_container_path
@@ -217,7 +218,7 @@
   Boolean $install_pam            = true,
 
   # Source install
-  String $version = '21.08.8',
+  String $version = '23.11.5',
   Array $source_dependencies = [],
   Array $configure_flags = [],
   Boolean $source_install_manage_alternatives = true,
@@ -240,6 +241,7 @@
   Boolean $reload_services                            = false,
   Boolean $restart_services                           = true,
   Integer $slurmctld_conn_validator_timeout           = 60,
+  Boolean $reconfig_ignore_errors = false,
 
   # User and group management
   Boolean $manage_slurm_user = true,
@@ -385,9 +387,8 @@
   # cgroups
   String $cgroup_conf_template             = 'slurm/cgroup/cgroup.conf.erb',
   Optional[String] $cgroup_conf_source               = undef,
-  Boolean $cgroup_automount                 = true,
   Stdlib::Absolutepath $cgroup_mountpoint                = '/sys/fs/cgroup',
-  Optional[String] $cgroup_plugin = undef,
+  String $cgroup_plugin = 'autodetect',
   Integer $cgroup_allowed_ram_space         = 100,
   Integer $cgroup_allowed_swap_space        = 0,
   Boolean $cgroup_constrain_cores           = false,
@@ -398,6 +399,7 @@
   Integer $cgroup_max_swap_percent          = 100,
   Optional[Integer[0,100]] $cgroup_memory_swappiness = undef,
   Integer $cgroup_min_ram_space             = 30,
+  Optional[Boolean] $cgroup_signal_child_processes = undef,
 
   # OCI
   String $oci_conf_template             = 'slurm/oci.conf.erb',

diff --git a/manifests/job_container.pp b/manifests/job_container.pp
@@ -4,25 +4,37 @@
 #   job_container.conf BasePath
 # @param auto_base_path
 #   job_container.conf AutoBasePath
+# @param dirs
+#   job_container.conf Dirs
 # @param init_script
 #   job_container.conf InitScript
 # @param node_name
 #   job_container.conf NodeName
+# @param shared
+#   job_container.conf Shared
 # @param order
 #   Order in job_container.conf
 #
 define slurm::job_container (
   Stdlib::Absolutepath $base_path,
   Boolean $auto_base_path                     = false,
+  Optional[Array[Stdlib::Absolutepath]] $dirs = undef,
   Optional[Stdlib::Absolutepath] $init_script = undef,
   Optional[String] $node_name                 = undef,
+  Optional[Boolean] $shared = undef,
   Variant[String[1], Integer] $order = '50',
 ) {
   include slurm
 
   $_base_path = "BasePath=${base_path}"
   $_auto_base_path = "AutoBasePath=${auto_base_path}"
 
+  if $dirs {
+    $_dirs = "Dirs=${dirs.join(',')}"
+  } else {
+    $_dirs = undef
+  }
+
   if $init_script {
     $_init_script = "InitScript=${init_script}"
   } else {
@@ -35,17 +47,30 @@
     $node_param = undef
   }
 
-  $params = [
-    $node_param,
-    $_auto_base_path,
-    $_base_path,
-    $_init_script,
-  ].filter |$p| { $p =~ NotUndef }
+  if $shared !~ Undef {
+    $shared_param = "Shared=${shared}"
+  } else {
+    $shared_param = undef
+  }
 
   if $node_name {
+    $params = [
+      $node_param,
+      $_auto_base_path,
+      $_base_path,
+      $_dirs,
+      $_init_script,
+      $shared_param,
+    ].filter |$p| { $p =~ NotUndef }
     $content = "${strip(join($params, ' '))}\n"
   } else {
-    $content = join($params, "\n")
+    $params = [
+      $_base_path,
+      $_dirs,
+      $_init_script,
+      $shared_param,
+    ].filter |$p| { $p =~ NotUndef }
+    $content = "${_auto_base_path}\n${params.join(' ')}"
   }
 
   concat::fragment { "job_container.conf-${name}":

diff --git a/manifests/params.pp b/manifests/params.pp
@@ -32,7 +32,7 @@
     'CliFilterPlugins' => undef,
     'CommunicationParameters' => undef,
     'CompleteWait' => '0',
-    'CoreSpecPlugin' => 'core_spec/none',
+    'CoreSpecPlugin' => undef,
     'CpuFreqDef' => undef,
     'CpuFreqGovernors' => 'OnDemand,Performance,UserSpace',
     'CredType' => 'cred/munge',
@@ -46,8 +46,6 @@
     'EioTimeout' => '60',
     'EnforcePartLimits' => 'NO',
     'EpilogMsgTime' => '2000',
-    'ExtSensorsFreq' => undef,
-    'ExtSensorsType' => undef,
     'FairShareDampeningFactor' => '1',
     'FederationParameters' => undef,
     'FirstJobId' => '1',
@@ -63,7 +61,7 @@
     'JobAcctGatherType' => 'jobacct_gather/cgroup',
     'JobAcctGatherFrequency' => 'task=30,energy=0,network=0,filesystem=0',
     'JobAcctGatherParams' => undef,
-    'JobCompType' => 'jobcomp/none',
+    'JobCompType' => undef,
     'JobContainerType' => undef,
     'JobFileAppend' => undef,
     'JobRequeue' => '1',
@@ -107,7 +105,7 @@
     'PriorityParameters' => undef,
     'PrioritySiteFactorParameters' => undef,
     'PrioritySiteFactorPlugin' => 'site_factor/none',
-    'PriorityType' => 'priority/basic',
+    'PriorityType' => 'priority/multifactor',
     'PriorityUsageResetPeriod' => 'NONE',
     'PriorityWeightAge' => '0',
     'PriorityWeightAssoc' => '0',
@@ -138,7 +136,7 @@
     'SchedulerTimeSlice' => '30',
     'SchedulerType' => 'sched/backfill',
     'ScronParameters' => undef,
-    'SelectType' => 'select/linear',
+    'SelectType' => 'select/cons_tres',
     'SelectTypeParameters' => undef,
     'SlurmctldAddr' => undef,
     'SlurmctldDebug' => 'info',
@@ -162,15 +160,15 @@
     'SuspendRate' => '60',
     'SuspendTime' => '-1',
     'SuspendTimeout' => '30',
-    'SwitchType' => 'switch/none',
+    'SwitchType' => undef,
     'TaskPlugin' => 'task/affinity,task/cgroup',
     'TaskPluginParam' => undef,
     'TCPTimeout' => '2',
     'TmpFS' => '/tmp',
     'TopologyParam' => undef,
-    'TopologyPlugin'  => 'topology/none',
+    'TopologyPlugin'  => undef,
     'TrackWCKey' => undef,
-    'TreeWidth' => '50',
+    'TreeWidth' => '16',
     'UnkillableStepProgram' => undef,
     'UnkillableStepTimeout' => '60',
     'UsePAM' => '0',

diff --git a/manifests/slurmctld/service.pp b/manifests/slurmctld/service.pp
@@ -85,7 +85,14 @@
     hasrestart => true,
   }
 
+  if $slurm::reconfig_ignore_errors {
+    $reconfig_command = 'scontrol reconfig || exit 0'
+  } else {
+    $reconfig_command = 'scontrol reconfig'
+  }
+
   exec { 'scontrol reconfig':
+    command     => $reconfig_command,
     path        => '/usr/bin:/bin:/usr/sbin:/sbin',
     refreshonly => true,
   }

diff --git a/metadata.json b/metadata.json
@@ -42,7 +42,7 @@
     },
     {
       "name": "puppet/systemd",
-      "version_requirement": ">= 3.1.0 <6.0.0"
+      "version_requirement": ">= 3.1.0 <7.0.0"
     },
     {
       "name": "puppet/archive",

diff --git a/spec/acceptance/01_slurmd_spec.rb b/spec/acceptance/01_slurmd_spec.rb
@@ -3,11 +3,7 @@
 require 'spec_helper_acceptance'
 
 describe 'slurmd' do
-  if ['docker', 'hyperv'].include?(fact('virtual'))
-    let(:slurm_user) { 'root' }
-  else
-    let(:slurm_user) { 'slurm' }
-  end
+  let(:slurm_user) { 'slurm' }
 
   context 'with default parameters' do
     nodes = hosts_as('slurmd')

diff --git a/spec/acceptance/02_slurmdbd_spec.rb b/spec/acceptance/02_slurmdbd_spec.rb
@@ -3,11 +3,7 @@
 require 'spec_helper_acceptance'
 
 describe 'slurmdbd' do
-  if ['docker', 'hyperv'].include?(fact('virtual'))
-    let(:slurm_user) { 'root' }
-  else
-    let(:slurm_user) { 'slurm' }
-  end
+  let(:slurm_user) { 'slurm' }
 
   context 'with default parameters' do
     nodes = hosts_as('slurmdbd')

diff --git a/spec/acceptance/03_slurmctld_spec.rb b/spec/acceptance/03_slurmctld_spec.rb
@@ -3,11 +3,7 @@
 require 'spec_helper_acceptance'
 
 describe 'slurmctld' do
-  if ['docker', 'hyperv'].include?(fact('virtual'))
-    let(:slurm_user) { 'root' }
-  else
-    let(:slurm_user) { 'slurm' }
-  end
+  let(:slurm_user) { 'slurm' }
 
   context 'with default parameters' do
     nodes = hosts_as('slurmctld')

diff --git a/spec/acceptance/05_slurmrestd_spec.rb b/spec/acceptance/05_slurmrestd_spec.rb
@@ -3,11 +3,7 @@
 require 'spec_helper_acceptance'
 
 describe 'slurmrestd' do
-  if ['docker', 'hyperv'].include?(fact('virtual'))
-    let(:slurm_user) { 'root' }
-  else
-    let(:slurm_user) { 'slurm' }
-  end
+  let(:slurm_user) { 'slurm' }
 
   context 'with default parameters' do
     nodes = hosts_as('slurmdbd')

diff --git a/spec/defines/slurm_job_container_spec.rb b/spec/defines/slurm_job_container_spec.rb
@@ -11,15 +11,15 @@
                   ]).each do |_os, os_facts|
     let(:facts) { os_facts }
     let(:title) { '/dev/shm' }
-    let(:params) { { auto_base_path: true, base_path: '/dev/shm/slurm' } }
+    let(:params) { { auto_base_path: true, base_path: '/tmp', dirs: ['/dev/shm'], shared: true } }
 
     it { is_expected.to create_slurm__job_container('/dev/shm') }
     it { is_expected.to contain_class('slurm') }
 
     it do
       is_expected.to contain_concat__fragment('job_container.conf-/dev/shm').with(
         target: 'job_container.conf',
-        content: "AutoBasePath=true\nBasePath=/dev/shm/slurm",
+        content: "AutoBasePath=true\nBasePath=/tmp Dirs=/dev/shm Shared=true",
         order: '50',
       )
     end