Support for provisioning nvidia gpu agents

modules/agent_files/templates/xorg.conf.erb

@@ -0,0 +1,48 @@
+Section "ServerLayout"
+    Identifier     "Layout0"
+    InputDevice    "Keyboard0" "CoreKeyboard"
+    InputDevice    "Mouse0" "CorePointer"
+EndSection
+
+Section "Files"
+EndSection
+
+Section "InputDevice"
+    # generated from default
+    Identifier     "Mouse0"
+    Driver         "mouse"
+    Option         "Protocol" "auto"
+    Option         "Device" "/dev/psaux"
+    Option         "Emulate3Buttons" "no"
+    Option         "ZAxisMapping" "4 5"
+EndSection
+
+Section "InputDevice"
+    # generated from default
+    Identifier     "Keyboard0"
+    Driver         "kbd"
+EndSection
+
+Section "Monitor"
+    Identifier     "Monitor0"
+    VendorName     "Unknown"
+    ModelName      "Unknown"
+    HorizSync       28.0 - 33.0
+    VertRefresh     43.0 - 72.0
+    Option         "DPMS"
+EndSection
+
+Section "Device"
+    Identifier     "Device0"
+    Driver         "nvidia"
+    VendorName     "NVIDIA Corporation"
+    BoardName      "GRID K520"
+    BusID          "<%= @facts['gpu_device_bus_id'] %>"
+EndSection
+
+Section "Screen"
+    Identifier     "Default Screen"
+    Device         "Device0"
+    Monitor        "Monitor0"
+    Option         "AllowEmptyInitialConfiguration" "True"
+EndSection

modules/facts/lib/facter/busid.rb

@@ -0,0 +1,5 @@
+Facter.add(:gpu_device_bus_id) do
+  setcode do
+    Facter::Core::Execution.execute('nvidia-xconfig --query-gpu-info  | grep BusID | sed "s/.*PCI:/PCI:/g"')
+  end
+end

modules/profile/files/jenkins/agent_gpu/etc/lightdm/lightdm.conf

@@ -0,0 +1,2 @@
+[SeatDefaults]
+display-setup-script=/etc/lightdm/xhost.sh

modules/profile/files/jenkins/agent_gpu/etc/lightdm/xhost.sh

@@ -0,0 +1,3 @@
+#!/bin/sh
+xhost +si:localuser:jenkins-agent
+touch /tmp/xhost_`date +"%T"`

modules/profile/files/jenkins/agent_gpu/nvidia-docker.list

@@ -0,0 +1,3 @@
+deb https://nvidia.github.io/libnvidia-container/ubuntu16.04/$(ARCH) /
+deb https://nvidia.github.io/nvidia-container-runtime/ubuntu16.04/$(ARCH) /
+deb https://nvidia.github.io/nvidia-docker/ubuntu16.04/$(ARCH) /

modules/profile/manifests/jenkins/agent_gpu.pp

@@ -0,0 +1,101 @@
+# Jenkins Agent Profile
+#
+# Profile class for a node configured to act as a swarm agent for Jenkins.
+# This profile should only ever be declared with an include into a role or site manifest.
+# Parameter overloading should be done using hiera automatic parameter lookup.
+#
+# @example
+#    include profile::jenkins::master
+#
+# @pararm agent_username The unix user the agent will configure and run as.
+class profile::jenkins::agent_gpu {
+
+  include apt
+
+  # neeed for xhost
+  package { 'x11-xserver-utils' :
+    ensure => installed,
+  }
+
+  if $facts['ec2_instance_id'] {
+    package { 'linux-aws':
+      ensure => installed,
+      # When running in EC2 the AWS kernel needs to be installed before
+      # compiling the nvidia driver.
+      # TODO(nuclearsandwich) Does the xorg.conf really depend on the kernel or
+      # is it implicit based on drivers?
+      before => [ File['/etc/X11/xorg.conf'], Package['nvidia-375'] ]
+    }
+  }
+
+  package { 'xserver-xorg-dev':
+    ensure => installed,
+  }
+
+  # needs to update first the kernel and headers before
+  # compiling the nvidia driver
+  package { 'nvidia-375':
+    ensure  => installed,
+  }
+
+  file { '/etc/X11/xorg.conf':
+    content => template('agent_files/xorg.conf.erb'),
+    mode    => '0744',
+    require => [
+      Package[lightdm],
+      Package['nvidia-375'],
+      Package['x11-xserver-utils'],
+      Package[xserver-xorg-dev],
+    ],
+  }
+
+  apt::key { 'nvidia_docker_key' :
+    source => 'https://nvidia.github.io/nvidia-docker/gpgkey',
+    id     => 'C95B321B61E88C1809C4F759DDCAE044F796ECB0',
+  }
+
+  file { '/etc/apt/sources.list.d/nvidia-docker.list':
+    source  => 'puppet:///modules/profile/jenkins/agent_gpu/nvidia-docker.list',
+    require => Apt::Key['nvidia_docker_key'],
+    notify  =>  Exec['apt_update']
+  }
+
+  package { 'nvidia-docker2':
+    ensure  => installed,
+    require => File['/etc/apt/sources.list.d/nvidia-docker.list']
+  }
+
+  package { 'lightdm':
+    ensure => installed,
+  }
+
+  file { '/etc/lightdm/xhost.sh':
+    source  => 'puppet:///modules/profile/jenkins/agent_gpu/etc/lightdm/xhost.sh',
+    mode    => '0744',
+    require => [ Package[lightdm], Package[x11-xserver-utils] ]
+  }
+
+  # This two rules do: check if no lightdm is present and create one
+  # Ensure that display-setup-script is set
+
+  file { '/etc/lightdm/lightdm.conf':
+    ensure  => 'present',
+    source  => 'puppet:///modules/profile/jenkins/agent_gpu/etc/lightdm/lightdm.conf',
+    replace => 'no', # this is the important property
+    require => [ File['/etc/lightdm/xhost.sh'], File['/etc/X11/xorg.conf'] ]
+  }
+
+  file_line { '/etc/lightdm/lightdm.conf':
+    ensure  => present,
+    require => File['/etc/lightdm/lightdm.conf'],
+    line    => 'display-setup-script=/etc/lightdm/xhost.sh',
+    notify  => Service[lightdm],
+    path    => '/etc/lightdm/lightdm.conf',
+  }
+
+  service { 'lightdm':
+    ensure     => running,
+    enable     => true,
+    hasrestart => true,
+  }
+}

modules/profile/manifests/ros/base.pp

@@ -42,7 +42,7 @@
     $defaults = {
       'ensure' => 'present',
     }
-    create_resources(ssh_authorized_key, hiera('ssh_keys'), $defaults)
+    # create_resources(ssh_authorized_key, hiera('ssh_keys'), $defaults)
   }
   else{
     notice("No ssh_keys defined. You should probably have at least one.")

modules/role/manifests/buildfarm/agent_gpu.pp

@@ -0,0 +1,6 @@
+class role::buildfarm::agent_gpu {
+  # Find the other instances
+  include profile::ros::base
+  include profile::jenkins::agent
+  include profile::jenkins::agent_gpu
+}