ansible/playbooks/vhost_reboot.yml

# requires --extra-vars="target=somevhost fqdn"

#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

# TO BE DONE - should be fixable w/ansible 0.9
# wait for all of the instances to return(?)
# compare the first list to the second for state info
# how to do this:
# capture output of action: virt command=info before halting guests
# compare to same command after vhost comes back
# ansible 0.9 should allow us to preserve content of two registered variables
# across multiple plays


- name: find instances
  hosts: $target
  user: root

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

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


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

  tasks:
  - name: schedule host downtime
    action: nagios action=downtime minutes=30 service=host host={{ inventory_hostname_short }}
    delegate_to: noc01.phx2.fedoraproject.org
    ignore_errors: true
    when: not ansible_domain.startswith('stg')

  - name: schedule host downtime
    action: nagios action=downtime minutes=30 service=host host={{ inventory_hostname_short }}.stg
    delegate_to: noc01.phx2.fedoraproject.org
    ignore_errors: true
    when: ansible_domain.startswith('stg')

  - name: halt the instances - to poweroff
    action: command /sbin/halt -p
    ignore_errors: true
    # if one of them is down we don't care


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

  tasks:
  - name: tell nagios to shush
    action: nagios action=downtime minutes=60 service=host host=$inventory_hostname
    delegate_to: noc01.phx2.fedoraproject.org
    ignore_errors: true

  - name: halt -r the $target
    action: command /sbin/reboot

  - name: wait for $target to come back - up to 6 minutes
    local_action: wait_for host=$target port=22 delay=120 timeout=420

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

  - name: sync time
    command: ntpdate -u 66.187.233.4