puppet-slurm_providers

Table of Contents

1. Overview
   • Supported versions of SLURM
2. Setup - The basics of getting started
3. Reference - An under-the-hood peek at what the module is doing
4. Limitations - OS compatibility, etc.

Overview

The SLURM providers module lets you manage various SLURM resources with Puppet.

Supported versions of SLURM

Currenlty this module supports version 19.05 of SLURM

Setup

This module requires that sacctmgr be in PATH.

If SLURM binaries are is not in path then then you must configure Puppet with a valid path to sacctmgr and scontrol. Below is an example of configuring Puppet if SLURM install prefix is /opt/slurm.

slurm_config { 'puppet':
  sacctmgr_path => '/opt/slurm/bin/sacctmgr',
  scontrol_path => '/opt/slurm/bin/scontrol',
}

Reference

http://treydock.github.io/puppet-slurm_providers/

Limitations

This module has been tested using the following versions of SLURM

• 20.02.x
• 20.11.x
• 21.08.x
• 22.05.x
• 23.02.x
• 23.11.x

The following operating systems have been tested

• RHEL/CentOS 7 x86_64
• RHEL/Rocky 8 x86_64
• RHEL/Rocky 9 x86_64

Development

Testing

Install gem dependencies

bundle install

Run unit tests

bundle exec rake spec

The following environment variables can be used to modify the behavior of the beaker tests:

• SLURM_BEAKER_version - Version of SLURM to install. Defaults to 20.02.3

Example of running beaker tests using an internal repository, and leaving VMs running after the tests.

export BEAKER_destroy=no
export BEAKER_PUPPET_COLLECTION=puppet5
export PUPPET_INSTALL_TYPE=agent
export BEAKER_set=centos-7
bundle exec rake beaker

Reference

Table of Contents

Resource types

• slurm_account: Puppet type that manages a SLURM account
• slurm_cluster: Puppet type that manages a SLURM cluster
• slurm_config: Abstract type to configure other SLURM types
• slurm_license: Puppet type that manages a SLURM software resource
• slurm_qos: Puppet type that manages a SLURM QOS
• slurm_reservation: Puppet type that manages a SLURM Reservation
• slurm_user: Puppet type that manages a SLURM user
• slurmctld_conn_validator: Verify that a connection can be successfully established between a node and the slurmctld server. Its primary use is as a precondition to pr
• slurmdbd_conn_validator: Verify that a connection can be successfully established between a node and the slurmdbd server. Its primary use is as a precondition to pre

Resource types

slurm_account

Puppet type that manages a SLURM account

Examples

Add SLURM account

slurm_account { 'staff on cluster':
  ensure    => 'present',
  max_jobs  => 1000,
  priority  => 9999,
}

@example Add SLURM account
  slurm_account { 'staff:cluster':
    ensure    => 'present',
    max_jobs  => 1000,
    priority  => 9999,
  }

Properties

The following properties are available in the slurm_account type.

default_qos

DefaultQOS

description

Description

ensure

Valid values: present, absent

The basic property that the resource should be in.

Default value: present

fairshare

Fairshare number or "parent"

Default value: 1

grp_jobs

GrpJobs

Default value: absent

grp_jobs_accrue

GrpJobsAccrue

Default value: absent

grp_submit_jobs

GrpSubmitJobs

Default value: absent

grp_tres

GrpTRES

Default value: absent

grp_tres_mins

GrpTRESMins

Default value: absent

grp_tres_run_mins

GrpTRESRunMins

Default value: absent

grp_wall

GrpWall

Default value: absent

max_jobs

MaxJobs

Default value: absent

max_jobs_accrue

MaxJobsAccrue

Default value: absent

max_submit_jobs

MaxSubmitJobs

Default value: absent

max_tres_mins_per_job

MaxTresMinsPerJob

Default value: absent

max_tres_per_job

MaxTresPerJob

Default value: absent

max_tres_per_node

MaxTresPerJob

Default value: absent

max_wall_duration_per_job

MaxWallDurationPerJob

Default value: absent

organization

Organization

parent_name

Account parent name

priority

Priority

qos

QOS, undefined will inherit parent QOS

Parameters

The following parameters are available in the slurm_account type.

• account
• cluster
• name
• provider

account

Account name

cluster

Cluster name

name

namevar

Account name

provider

The specific backend to use for this slurm_account resource. You will seldom need to specify this --- Puppet will usually discover the appropriate provider for your platform.

slurm_cluster

Puppet type that manages a SLURM cluster

Examples

Add a SLURM cluster

slurm_cluster { 'test':
  ensure => 'present',
}

Properties

The following properties are available in the slurm_cluster type.

ensure

Valid values: present, absent

The basic property that the resource should be in.

Default value: present

features

Features

Default value: absent

fed_state

Valid values: active, inactive, drain, drain_remove

FedState

federation

Federation

Default value: absent

Parameters

The following parameters are available in the slurm_cluster type.

• flags
• name
• provider

flags

Flags

name

namevar

cluster name

provider

The specific backend to use for this slurm_cluster resource. You will seldom need to specify this --- Puppet will usually discover the appropriate provider for your platform.

slurm_config

Abstract type to configure other SLURM types

Parameters

The following parameters are available in the slurm_config type.

• install_prefix
• name

install_prefix

The path to SLURM install prefix

name

namevar

The name of the resource

slurm_license

Puppet type that manages a SLURM software resource

Examples

Add SLURM software resource

slurm_license { 'matlab@host':
  ensure  => 'present',
  count   => 100,
}
slurm_license { 'matlab@host for linux':
  ensure          => 'present',
  percent_allowed => 100,
}

Properties

The following properties are available in the slurm_license type.

count

Count

description

Description

ensure

Valid values: present, absent

The basic property that the resource should be in.

Default value: present

percent_allowed

PercentAllowed

server_type

ServerType

Default value: absent

Parameters

The following parameters are available in the slurm_license type.

• cluster
• name
• provider
• resource_name
• server
• type

cluster

Cluster

name

namevar

Resource name

provider

The specific backend to use for this slurm_license resource. You will seldom need to specify this --- Puppet will usually discover the appropriate provider for your platform.

resource_name

Resource name

server

Server

type

Resource type, read-only

Default value: License

slurm_qos

Puppet type that manages a SLURM QOS

Examples

Add SLURM QOS

slurm_qos { 'high':
  ensure            => 'present',
  flags             => ['DenyOnLimit','RequiresReservation'],
  grace_time        => 300,
  grp_tres          => { 'node' => 40 },
  max_tres_per_user => { 'node' => 20 },
  max_wall          => '2-00:00:00',
  priority          => 2000000,
}

Properties

The following properties are available in the slurm_qos type.

description

Description

ensure

Valid values: present, absent

The basic property that the resource should be in.

Default value: present

flags

Valid values: DenyOnLimit, EnforceUsageThreshold, NoReserve, PartitionMaxNodes, PartitionMinNodes, OverPartQOS, PartitionTimeLimit, RequiresReservation, NoDecay, UsageFactorSafe, absent

Flags

Default value: absent

grace_time

GraceTime

Default value: 0

grp_jobs

GrpJobs

Default value: absent

grp_jobs_accrue

GrpJobsAccrue

Default value: absent

grp_submit_jobs

GrpSubmitJobs

Default value: absent

grp_tres

GrpTRES

Default value: absent

grp_tres_mins

GrpTRESMins

Default value: absent

grp_tres_run_mins

GrpTRESRunMins

Default value: absent

grp_wall

GrpWall

Default value: absent

max_jobs_per_account

MaxJobsPerAccount

Default value: absent

max_jobs_per_user

MaxJobsPerUser

Default value: absent

max_submit_jobs_per_account

MaxSubmitJobsPerAccount

Default value: absent

max_submit_jobs_per_user

MaxSubmitJobsPerUser

Default value: absent

max_tres_mins

MaxTresMins

Default value: absent

max_tres_per_account

MaxTresPerAccount

Default value: absent

max_tres_per_job

MaxTresPerJob

Default value: absent

max_tres_per_node

MaxTresPerNode

Default value: absent

max_tres_per_user

MaxTresPerUser

Default value: absent

max_wall

MaxWall

Default value: absent

min_prio_threshold

MinPrioThreshold

Default value: absent

min_tres_per_job

MinTRESPerJob

Default value: absent

preempt

Preempt

Default value: absent

preempt_exempt_time

PreemptExemptTime

Default value: absent

preempt_mode

Valid values: cluster, cancel, checkpoint, requeue

PreemptMode

Default value: cluster

priority

Priority

Default value: 0

usage_factor

UsageFactor

Default value: 1.000000

usage_threshold

UsageThreshold

Default value: absent

Parameters

The following parameters are available in the slurm_qos type.

• name
• provider

name

namevar

QOS name

provider

The specific backend to use for this slurm_qos resource. You will seldom need to specify this --- Puppet will usually discover the appropriate provider for your platform.

slurm_reservation

Puppet type that manages a SLURM Reservation

Examples

Add SLURM Reservation

slurm_reservation { 'maint':
  ensure     => 'present',
  start_time => 'now',
  duration   => '02:00:00',
  users      => ['root'],
  flags      => ['maint','ignore_jobs'],
  nodes      => 'ALL',
}

Properties

The following properties are available in the slurm_reservation type.

accounts

Accounts

burst_buffer

BurstBuffer

core_cnt

CoreCnt

duration

Duration

end_time

EndTime

ensure

Valid values: present, absent

The basic property that the resource should be in.

Default value: present

flags

Flags

licenses

Licenses

node_cnt

NodeCnt

nodes

Nodes

partition_name

PartitionName

start_time

StartTime

tres

TRES

users

Users

Parameters

The following parameters are available in the slurm_reservation type.

• features
• name
• provider
• timezone

features

Features

name

namevar

Reservation name

provider

The specific backend to use for this slurm_reservation resource. You will seldom need to specify this --- Puppet will usually discover the appropriate provider for your platform.

timezone

TZ environment variable value

slurm_user

Puppet type that manages a SLURM user

Examples

Add SLURM user under account 'bar' on cluster 'test'

slurm_user { 'foo under bar on test':
  ensure    => 'present',
  max_jobs  => 1000,
  priority  => 9999,
}

Add SLURM user under account 'bar' on cluster 'test'

slurm_user { 'foo:bar:test':
  ensure    => 'present',
  max_jobs  => 1000,
  priority  => 9999,
}

Properties

The following properties are available in the slurm_user type.

admin_level

Valid values: None, Operator, Administrator

AdminLevel

Default value: None

default_account

DefaultAccount

default_qos

DefaultQOS

ensure

Valid values: present, absent

The basic property that the resource should be in.

Default value: present

fairshare

Fairshare number or "parent"

Default value: 1

grp_jobs

GrpJobs

Default value: absent

grp_jobs_accrue

GrpJobsAccrue

Default value: absent

grp_submit_jobs

GrpSubmitJobs

Default value: absent

grp_tres

GrpTRES

Default value: absent

grp_tres_mins

GrpTRESMins

Default value: absent

grp_tres_run_mins

GrpTRESRunMins

Default value: absent

grp_wall

GrpWall

Default value: absent

max_jobs

MaxJobs

Default value: absent

max_jobs_accrue

MaxJobsAccrue

Default value: absent

max_submit_jobs

MaxSubmitJobs

Default value: absent

max_tres_mins_per_job

MaxTresMinsPerJob

Default value: absent

max_tres_per_job

MaxTresPerJob

Default value: absent

max_tres_per_node

MaxTresPerJob

Default value: absent

max_wall_duration_per_job

MaxWallDurationPerJob

Default value: absent

priority

Priority

qos

QOS, undefined will inherit parent QOS

Parameters

The following parameters are available in the slurm_user type.

• account
• cluster
• name
• partition
• provider
• user

account

Account name

cluster

Cluster name

name

namevar

User name

partition

Partition name

Default value: absent

provider

The specific backend to use for this slurm_user resource. You will seldom need to specify this --- Puppet will usually discover the appropriate provider for your platform.

user

User name

slurmctld_conn_validator

Verify that a connection can be successfully established between a node and the slurmctld server. Its primary use is as a precondition to prevent configuration changes from being applied if the slurmctld server cannot be reached.

Properties

The following properties are available in the slurmctld_conn_validator type.

ensure

Valid values: present, absent

The basic property that the resource should be in.

Default value: present

Parameters

The following parameters are available in the slurmctld_conn_validator type.

• name
• provider
• timeout

name

namevar

An arbitrary name used as the identity of the resource.

provider

The specific backend to use for this slurmctld_conn_validator resource. You will seldom need to specify this --- Puppet will usually discover the appropriate provider for your platform.

timeout

The max number of seconds that the validator should wait before giving up and deciding that slurmctld is not running; defaults to 30 seconds.

Default value: 30

slurmdbd_conn_validator

Verify that a connection can be successfully established between a node and the slurmdbd server. Its primary use is as a precondition to prevent configuration changes from being applied if the slurmdbd server cannot be reached.

Properties

The following properties are available in the slurmdbd_conn_validator type.

ensure

Valid values: present, absent

The basic property that the resource should be in.

Default value: present

Parameters

The following parameters are available in the slurmdbd_conn_validator type.

• name
• provider
• timeout

name

namevar

An arbitrary name used as the identity of the resource.

provider

The specific backend to use for this slurmdbd_conn_validator resource. You will seldom need to specify this --- Puppet will usually discover the appropriate provider for your platform.

timeout

The max number of seconds that the validator should wait before giving up and deciding that slurmdbd is not running; defaults to 30 seconds.

Default value: 30

Change log

All notable changes to this project will be documented in this file. The format is based on Keep a Changelog and this project adheres to Semantic Versioning.

v0.14.1 (2024-03-25)

Full Changelog

Fixed

• Support Slurm 23.11 #36 (treydock)

v0.14.0 (2023-12-28)

Full Changelog

Changed

• MAJOR UPDATES: Read description #34 (treydock)

Added

• Support Slurm 23.02.x #35 (treydock)

v0.13.0 (2022-10-06)

Full Changelog

Added

• Support hourly reservations #33 (treydock)

v0.12.4 (2022-08-08)

Full Changelog

Fixed

• Fix parsing of reservation with no flags #32 (treydock)

v0.12.3 (2021-10-05)

Full Changelog

Fixed

• Fix slurm_qos having preempt changed to absent #31 (treydock)

v0.12.2 (2021-07-28)

Full Changelog

Fixed

• slurm_qos will autorequire QOS it preempts #30 (treydock)

v0.12.1 (2021-07-21)

Full Changelog

Fixed

• Improved handling of setting properties to absent #29 (treydock)

v0.12.0 (2021-03-17)

Full Changelog

Changed

• Drop Puppet 5 support, add Puppet 7 #28 (treydock)

v0.11.1 (2021-01-29)

Full Changelog

Fixed

• Improved property validation for slurm_license #27 (treydock)

v0.11.0 (2021-01-12)

Full Changelog

Added

• Support SLURM 20.11 #25 (treydock)

Fixed

• Fix slurmdbd_conn_validator to not exit early on error #26 (treydock)

v0.10.0 (2021-01-04)

Full Changelog

Added

• Support user association by partition #23 (treydock)

v0.9.0 (2020-10-09)

Full Changelog

Added

• Improved error handling #22 (treydock)

v0.8.0 (2020-09-16)

Full Changelog

Added

• Support timezone parameter for slurm_reservation #21 (treydock)

Fixed

• Fix autorequires for slurm_reservation #20 (treydock)

v0.7.3 (2020-09-15)

Full Changelog

Fixed

• Numerous bug fixes for slurm_reservation #19 (treydock)

v0.7.2 (2020-08-26)

Full Changelog

Fixed

• Fix handling of slurm_qos description to always be lower case #18 (treydock)

v0.7.1 (2020-07-29)

Full Changelog

Fixed

• Properly handle changes to AdminLevel for users #17 (treydock)

v0.7.0 (2020-07-22)

Full Changelog

Added

• Add slurmctld_conn_validator type #16 (treydock)

v0.6.1 (2020-07-13)

Full Changelog

Fixed

• Fix for slurmdbd_conn_validator #15 (treydock)

v0.6.0 (2020-07-13)

Full Changelog

Added

• Add slurmdbd_conn_validator type #14 (treydock)
• Add slurm_license type/provider #13 (treydock)

v0.5.0 (2020-06-23)

Full Changelog

Added

• Add slurm_user type and some misc fixes for other types #12 (treydock)

v0.4.2 (2020-06-22)

Full Changelog

Fixed

• Remove defaults for slurm_account for default_qos and priority #11 (treydock)

v0.4.1 (2020-06-22)

Full Changelog

Fixed

• Remove default from slurm_account description and organization #10 (treydock)

v0.4.0 (2020-06-22)

Full Changelog

Added

• Support more condensed composite names for slurm_account #9 (treydock)

v0.3.0 (2020-06-22)

Full Changelog

Added

• Improvements to slurm_account type #8 (treydock)
• Add slurm_account type/provider #6 (treydock)

v0.2.0 (2019-12-27)

Full Changelog

Changed

• Change slurm_config to only need install_prefix parameter #4 (treydock)

Added

• Add support for managing reservations #3 (treydock)

Fixed

• Fix acceptance tests #5 (treydock)

v0.1.1 (2019-12-23)

Full Changelog

Fixed

• Fix how sacctmgr default path is detected #2 (treydock)

v0.1.0 (2019-12-23)

Full Changelog

Changed

• Full rewrite of module #1 (treydock)

* This Changelog was automatically generated by github_changelog_generator