WIP slurm compute node image

ansible/linux-slurmcompute.yml

@@ -0,0 +1,6 @@
+---
+
+- hosts: all
+  become: yes
+  roles:
+    - jupyter-repo2docker

ansible/roles/linux-slurmcompute/defaults/main.yml

@@ -0,0 +1,9 @@
+openhpc_slurm_packages:
+  - ohpc-base-compute
+  - ohpc-slurm-client
+  - munge
+openhpc_extra_packages:
+  - lmod-ohpc
+  - slurm-libpmi-ohpc # to allow intel mpi to work properly
+  - ohpc-gnu9-openmpi4-perf-tools # for hpctests
+  - openblas-gnu9-ohpc # for hpctests (HPL)

ansible/roles/linux-slurmcompute/tasks/bootstrap.yml

@@ -0,0 +1,32 @@
+- name: Move rocky user
+  ansible.builtin.user:
+    name: rocky
+    home: /var/lib/rocky
+    move_home: true
+    local: true
+  become_method: "sudo"
+  # Need to change working directory otherwise we try to switch back to non-existent directory.
+  become_flags: '-i'
+
+- name: Reset ssh connection to allow user changes to affect ansible_user
+  meta: reset_connection
+
+- name: Set SELinux state and policy
+  ansible.posix.selinux:
+    state: permissive
+    policy: targeted
+
+- name: Update base image packages
+  ansible.builtin.dnf:
+    name: '*'
+    state: 'latest'
+  async: "{{ 30 * 60 }}" # wait for up to 30 minutes
+  poll: 15 # check every 15 seconds
+
+- name: Reboot to cope with package updates and SELinux changes
+  ansible.builtin.reboot:
+    post_reboot_delay: 30
+
+- name: Wait for hosts to be reachable
+  ansible.builtin.wait_for_connection:
+    sleep: 15

ansible/roles/linux-slurmcompute/tasks/main.yml

@@ -0,0 +1,25 @@
+- name: Get builder commit
+  shell:
+    cmd: git describe --all --long --dirty
+  register: builder_commit
+  delegate_to: localhost
+
+- name: Write builder commit to /var/lib/misc/build.txt
+  copy:
+    dest: /var/lib/misc/build.txt
+    content: "{{ builder_commit.stdout }}"
+
+- include_role:
+    name: linux-ansible-init
+
+- import_tasks: bootstrap.yml
+#- import_tasks: reimage.yml #  TODO: if required to support slurm-driven rebuild from compute nodes
+- import_tasks: slurm.yml
+- import_tasks: ood_vnc.yml
+- import_tasks: monitoring.yml
+
+- name: Delete /etc/resolv.conf
+  # required as if cloud-init (rather than network manager) controls this on next boot it won't be entirely overrwritten
+  file:
+    path: /etc/resolv.conf
+    state: absent

ansible/roles/linux-slurmcompute/tasks/monitoring.yml

@@ -0,0 +1,3 @@
+- name: Deploy node_exporter
+  import_role:
+    name: cloudalchemy.node_exporter

ansible/roles/linux-slurmcompute/tasks/ood_jupyter.yml

@@ -0,0 +1,16 @@
+# See https://osc.github.io/ood-documentation/latest/app-development/tutorials-interactive-apps/add-jupyter/software-requirements.html
+# - Will already have openssl and lmod
+
+- name: Install jupyter venv
+  # Requires separate step so that the upgraded pip is used to install packages
+  ansible.builtin.pip:
+    name: pip
+    state: latest
+    virtualenv: /opt/jupyter
+    virtualenv_command: python3 -m venv
+
+- name: Install jupyter package in venv
+  ansible.builtin.pip:
+    name: jupyter
+    virtualenv: /opt/jupyter
+    virtualenv_command: python3 -m venv

ansible/roles/linux-slurmcompute/tasks/ood_vnc.yml

@@ -0,0 +1,45 @@
+- name: Enable TurboVNC repo
+  ansible.builtin.get_url:
+    url: https://turbovnc.org/pmwiki/uploads/Downloads/TurboVNC.repo
+    dest: /etc/yum.repos.d/TurboVNC.repo
+
+- name: Install EPEL
+  ansible.builtin.dnf:
+    name: epel-release
+
+- name: Install VNC-related packages
+  ansible.builtin.dnf:
+  name:
+    - turbovnc
+    - nmap-ncat
+    - python3
+
+- name: Install Xfce desktop
+  ansible.builtin.dnf:
+    name: '@Xfce'
+
+- name: Install websockify venv
+  # Requires separate step so that the upgraded pip is used to install packages
+  ansible.builtin.pip:
+    name: pip
+    virtualenv: /opt/websockify
+    virtualenv_command: python3 -m venv
+  tags: install
+
+- name: Install websockify package in venv
+  ansible.builtin.pip:
+    name: websockify
+    virtualenv: /opt/websockify
+    virtualenv_command: python3 -m venv
+  tags: install
+
+- name: Symlink websockify to where Open Ondemand expects
+  ansible.builtin.file:
+    src: /opt/websockify/bin/websockify
+    dest: /opt/websockify/run
+    state: link
+
+- name: Disable screensaver # as users might not have passwords
+  ansible.builtin.dnf:
+    name: xfce4-screensaver
+    state: absent

ansible/roles/linux-slurmcompute/tasks/slurm.yml

@@ -0,0 +1,23 @@
+- name: Install ohpc-release package
+  yum:
+    name: "http://repos.openhpc.community/OpenHPC/2/CentOS_8/x86_64/ohpc-release-2-1.el8.x86_64.rpm"
+    state: present
+    disable_gpg_check: True
+
+- name: Enable PowerTools repo
+  community.general.ini_file:
+    path: /etc/yum.repos.d/CentOS-PowerTools.repo 8.3
+    create: false
+    option: enabled
+    value: 1
+
+- name: Install packages
+  ansible.builtin.dnf:
+    name: "{{ openhpc_slurm_packages + openhpc_extra_packages + }}"
+
+# - munge key
+# -  JobComp logfile exists?
+# - name: Set slurmctld location for configless operation
+#- name: Ensure Slurm service state
+# make sure munge and slurm wait for cloud-init
+

packer/linux-slurmcompute.pkr.hcl

@@ -0,0 +1,82 @@
+# Use like:
+#  $ PACKER_LOG=1 packer build --on-error=ask -var-file=<something>.pkrvars.hcl <thisfile>
+
+# "timestamp" template function replacement:s
+locals { timestamp = formatdate("YYMMDD-hhmm", timestamp())}
+
+variable "source_image_name" {
+  type = string
+}
+
+variable "network" {
+  type = string
+}
+
+variable "floating_ip_network" {
+  type = string
+}
+
+variable "flavor" {
+  type = string
+}
+
+variable "security_groups" {
+  type = list(string)
+}
+
+variable "volume_size" {
+  type = number
+  default = 10
+}
+
+variable "disk_format" {
+  type = string
+  default = "qcow2"
+}
+
+variable "distro_name" {
+  type = string
+}
+
+variable "ssh_username" {
+  type = string
+}
+
+source "openstack" "linux-slurmcompute" {
+  image_name = "${var.distro_name}-desktop-${local.timestamp}"
+  image_visibility = "private"
+  image_disk_format = "${var.disk_format}"
+
+  source_image_name = "${var.source_image_name}"
+  flavor = "${var.flavor}"
+  networks = ["${var.network}"]
+  security_groups = "${var.security_groups}"
+  floating_ip_network = "${var.floating_ip_network}"
+
+  use_blockstorage_volume = true
+  volume_size = "${var.volume_size}"
+
+  communicator = "ssh"
+  ssh_username = "${var.ssh_username}"
+  ssh_clear_authorized_keys = true
+}
+
+build {
+  source "source.openstack.linux-desktop" { }
+
+  provisioner "ansible" {
+    galaxy_file = "${path.root}/../requirements.yml"
+    playbook_file = "${path.root}/../ansible/linux-slurmcompute.yml"
+    use_proxy = false
+    extra_arguments = [
+      "-v",
+    ]
+    ansible_env_vars = ["ANSIBLE_SSH_RETRIES=10"]
+  }
+
+  post-processor "manifest" {
+    custom_data = {
+      source = "${source.name}"
+    }
+  }
+}