Add a playbook and role to configure Dell compute nodes to PXE boot

ansible/dell-compute-node-discovery.yml

@@ -0,0 +1,44 @@
+---
+- name: Ensure compute nodes are present in the Ansible inventory
+  hosts: config-mgmt
+  gather_facts: no
+  vars:
+    # Set this to a colon-separated list of compute node hostnames on which to
+    # trigger discovery. If unset, all compute nodes will be triggered.
+    compute_node_limit: ""
+    compute_node_limit_list: "{{ compute_node_limit.split(':') }}"
+  tasks:
+    - name: Add hosts for the compute nodes
+      add_host:
+        name: "{{ item.key }}"
+        groups: compute
+      with_dict: "{{ idrac_network_ips }}"
+      # Don't add hosts that already exist.
+      when:
+        - "{{ item.key not in groups['all'] }}"
+        - "{{ item.key | replace('-idrac', '') not in groups['all'] }}"
+        - "{{ not compute_node_limit or item.key | replace('-idrac', '') in compute_node_limit_list }}"
+      run_once: True
+
+- name: Ensure compute nodes are PXE booted
+  hosts: compute
+  gather_facts: no
+  vars:
+    # Set this to the index of the inteface on which to enable PXE.
+    drac_pxe_interface: 1
+    # Set this to a list of indices of all interfaces.
+    drac_pxe_all_interfaces: [1, 2, 3, 4]
+  tasks:
+    - name: Set a fact containing the compute node IPMI address
+      set_fact:
+        ansible_host: "{{ idrac_network_ips[inventory_hostname] }}"
+        ansible_user: "{{ ipmi_username }}"
+        ansible_ssh_pass: "{{ ipmi_password }}"
+
+    - name: Ensure the compute node's BMC is added as a known host
+      include_role:
+        role: ssh-known-host
+
+    - name: Ensure the compute node's BMC is set to PXE boot
+      include_role:
+        role: drac-pxe

ansible/roles/drac-pxe/defaults/main.yml

@@ -0,0 +1,15 @@
+---
+# Index of interface on which to enable PXE booting.
+drac_pxe_interface: 1
+
+# List of numerical indices of all interfaces.
+drac_pxe_all_interfaces: []
+
+# Timeout when waiting for boot sequence configuration to be applied.
+drac_pxe_timeout: 600
+
+# Number of times to attempt to perform write operations.
+drac_pxe_retries: 5
+
+# Interval between successive write operations.
+drac_pxe_interval: 5

ansible/roles/drac-pxe/tasks/main.yml

@@ -0,0 +1,97 @@
+---
+- name: Check the boot mode
+  raw: "racadm get BIOS.BiosBootSettings.BootMode"
+  register: result
+  failed_when: "'ERROR' in result.stdout"
+  changed_when: False
+
+- name: Set a fact containing the boot mode
+  set_fact:
+    # Format of last line is:
+    # BootMode=<boot mode>[ (Pending Value=<pending boot mode>)]
+    boot_mode_is_uefi: "{{ result.stdout_lines[-1] == 'BootMode=Uefi' }}"
+    boot_mode_is_bios: "{{ result.stdout_lines[-1] == 'BootMode=Bios' }}"
+
+- name: Fail if unable to determine the boot mode
+  fail:
+    msg: >
+      Unable to determine the boot mode. Got: {{ result.stdout }}. Expected
+      bios or uefi.
+  when:
+    - "{{ not boot_mode_is_bios }}"
+    - "{{ not boot_mode_is_uefi }}"
+
+- block:
+    - name: Ensure NIC boot protocol is configured
+      raw: "racadm set Nic.NICConfig.{{ item }}.LegacyBootProto {% if item == drac_pxe_interface %}PXE{% else %}NONE{% endif %}"
+      with_items: "{{ drac_pxe_all_interfaces }}"
+      register: result
+      failed_when: "'ERROR' in result.stdout"
+      until: "{{ drac_pxe_busy_message not in result.stdout }}"
+      retries: "{{ drac_pxe_retries }}"
+      delay: "{{ drac_pxe_interval }}"
+
+    - name: Ensure NIC configuration jobs are created
+      raw: "racadm jobqueue create NIC.Integrated.1-{{ item }}-1 -s TIME_NOW"
+      with_items: "{{ drac_pxe_all_interfaces }}"
+      register: result
+      failed_when: "'ERROR' in result.stdout"
+      until: "{{ drac_pxe_busy_message not in result.stdout }}"
+      retries: "{{ drac_pxe_retries }}"
+      delay: "{{ drac_pxe_interval }}"
+
+    - name: Ensure BIOS boot sequence is configured
+      raw: "racadm set BIOS.BiosBootSettings.bootseq NIC.Integrated.1-{{ drac_pxe_interface }}-1,HardDisk.List.1-1"
+      register: result
+      failed_when: "'ERROR' in result.stdout"
+      until: "{{ drac_pxe_busy_message not in result.stdout }}"
+      retries: "{{ drac_pxe_retries }}"
+      delay: "{{ drac_pxe_interval }}"
+
+  when: "{{ boot_mode_is_bios }}"
+
+- block:
+    - name: Ensure UEFI PXE device is configured
+      raw: "racadm set BIOS.PxeDev1Settings.PxeDev1Interface NIC.Integrated.1-{{ drac_pxe_interface }}-1"
+      register: result
+      failed_when: "'ERROR' in result.stdout"
+      until: "{{ drac_pxe_busy_message not in result.stdout }}"
+      retries: "{{ drac_pxe_retries }}"
+      delay: "{{ drac_pxe_interval }}"
+
+    - name: Ensure UEFI boot sequence is configured
+      raw: "racadm set BIOS.BiosBootSettings.UefiBootSeq NIC.PxeDevice.1-1,RAID.Integrated.1-1"
+      register: result
+      failed_when: "'ERROR' in result.stdout"
+      until: "{{ drac_pxe_busy_message not in result.stdout }}"
+      retries: "{{ drac_pxe_retries }}"
+      delay: "{{ drac_pxe_interval }}"
+
+  when: "{{ boot_mode_is_uefi }}"
+
+- name: Ensure BIOS configuration job is created
+  raw: "racadm jobqueue create BIOS.Setup.1-1 -s TIME_NOW"
+  register: result
+  failed_when: "'ERROR' in result.stdout"
+  until: "{{ drac_pxe_busy_message not in result.stdout }}"
+  retries: "{{ drac_pxe_retries }}"
+  delay: "{{ drac_pxe_interval }}"
+
+- name: Set a fact containing the BIOS configuration job ID
+  set_fact:
+    # Format of the last line is:
+    # JOB_ID = <job ID>
+    bios_job_id: "{{ result.stdout_lines[-1].split()[-1] }}"
+
+- name: Ensure server is rebooted
+  raw: "racadm serveraction powercycle"
+  register: result
+  failed_when: "'ERROR' in result.stdout"
+
+- name: Wait for the BIOS configuration job to complete
+  raw: "racadm jobqueue view -i {{ bios_job_id }}"
+  register: result
+  failed_when: "'ERROR' in result.stdout"
+  until: "{{ 'Status=Completed' in result.stdout }}"
+  retries: "{{ drac_pxe_timeout // drac_pxe_interval }}"
+  delay: "{{ drac_pxe_interval }}"

ansible/roles/drac-pxe/vars/main.yml

@@ -0,0 +1,3 @@
+---
+# Message emitted by the DRAC which can be retried.
+drac_pxe_busy_message: "ERROR: Lifecycle Controller is currently in use."