ansible/playbooks/vhost_reboot.yml

#
# This playbook lets you safely reboot a virthost and all it's guests.
#
# requires --extra-vars="target=somevhost fqdn"
# Might add nodns=true or nonagios=true to the extra vars

# General overview:
# talk to the vhost
# get back list of instances
# add each of their hostnames to an addhoc group
# halt each of them in a second play
# wait for them to die
# third play, reboot the vhost
#     wait for vhost to come back

# TODO: Figure out how to compare virt info pre and post boot.

---
- name: find instances
  vars_prompt:
  - name: target
    prompt: What is the target vhost (to reboot)
    private: false
  # Use a simple regexp to ignore spaces people might accidentally paste.
  hosts: "{{ target | regex_search('[-a-z0-9.]+')}}"
  gather_facts: false
  user: root

  tasks:
  - name: get list of guests
    virt: command=list_vms state=running
    register: vmlist

#  - name: get info on guests (prereboot)
#    virt: command=info
#    register: vminfo_pre

  - name: add them to myvms_new group
    local_action: add_host hostname={{ item }} groupname=myvms_new
    with_items: "{{ vmlist.list_vms }}"

  - name: add the vmhost to target group
    # Use a simple regexp to ignore spaces people might accidentally paste.
    local_action: add_host hostname={{ target | regex_search('[-a-z0-9.]+') }} groupname=target

# Call out to another playbook.  Disable any proxies that may live here
- name: update proxy dns if needed
  import_playbook: update-proxy-dns.yml
  vars:
    status: enable
    proxies: myvms_new:&proxies
  when: nodns is not defined or not "true" in nodns

- name: halt instances
  hosts: myvms_new
  user: root
  gather_facts: false
  serial: 1

  tasks:

  - name: schedule regular host downtime
    nagios: action=downtime minutes=30 service=host host={{ inventory_hostname_short }}{{ env_suffix }}
    delegate_to: noc01.iad2.fedoraproject.org
    ignore_errors: true
    when: nonagios is not defined or not nonagios

  - name: shutdown vms
    virt: command=shutdown name={{ inventory_hostname }}
    ignore_errors: true
    delegate_to: "{{ hostvars[groups['target'][0]]['ansible_fqdn'] }}"

- name: wait for the whole set to die.
  hosts: myvms_new
  gather_facts: false
  user: root

  tasks:
  - name: wait for them to die
    local_action: wait_for port=22 delay=30 timeout=300 state=stopped host={{ inventory_hostname }}

- name: reboot vhost
  hosts: "target"
  gather_facts: false
  user: root

  tasks:
  - name: tell nagios to shush
    nagios: action=downtime minutes=60 service=host host={{ inventory_hostname_short }}{{ env_suffix }}
    delegate_to: noc01.iad2.fedoraproject.org
    ignore_errors: true
    when: nonagios is not defined or not nonagios

  # With drive firmware and sometimes needed to poke things, make this 30m
  - name: reboot the virthost
    reboot: reboot_timeout=1800

  - name: wait for libvirtd to come back on the virthost
    wait_for: path=/run/libvirt/libvirt-sock state=present

  - name: look up vmlist
    virt: command=list_vms
    register: newvmlist

  - name: add them to myvms_postreboot group
    local_action: add_host hostname={{ item }} groupname=myvms_postreboot
    with_items: "{{ newvmlist.list_vms }}"

#  - name: sync time
#    ansible.builtin.command: ntpdate -u 1.rhel.pool.ntp.org

  - name: tell nagios to unshush
    nagios: action=unsilence service=host host={{ inventory_hostname_short }}{{ env_suffix }}
    delegate_to: noc01.iad2.fedoraproject.org
    ignore_errors: true
    when: nonagios is not defined or not nonagios

- name: post reboot tasks
  hosts: myvms_postreboot
  user: root
  gather_facts: false
  serial: 1

  tasks:

  - name: Wait for host to come back (ipa or rabbitmq)
    local_action: wait_for port=22 delay=30 timeout=300 state=started host={{ inventory_hostname }}
    when: inventory_hostname.startswith(('ipa', 'rabbitmq'))

  - name: restart gssproxy if we rebooted a ipa server
    service: name=gssproxy state=restarted
    when: inventory_hostname.startswith('ipa')

  - name: restart rabbitmq if we rebooted a rabbit server
    service: name=rabbitmq-server state=restarted
    when: inventory_hostname.startswith('rabbitmq')

# Call out to that dns playbook.  Put proxies back in now that they're back
- name: update proxy dns if needed
  import_playbook: update-proxy-dns.yml
  vars:
    status: enable
    proxies: myvms_new:&proxies
  when: nodns is not defined or not nodns

#  - name: get info on guests (postreboot)
#    virt: command=info
#    register: vminfo_post