ceph-ansible/infrastructure-playbooks/shrink-rgw.yml

---
# This playbook shrinks the Ceph RGW from your cluster
#
# Use it like this:
# ansible-playbook shrink-rgw.yml -e rgw_to_kill=ceph-rgw01
#     Prompts for confirmation to shrink, defaults to no and
#     doesn't shrink the cluster. yes shrinks the cluster.
#
# ansible-playbook -e ireallymeanit=yes|no shrink-rgw.yml
#     Overrides the prompt using -e option. Can be used in
#     automation scripts to avoid interactive prompt.


- name: confirm whether user really meant to remove rgw from the ceph cluster
  hosts: localhost
  become: false
  gather_facts: false
  vars_prompt:
    - name: ireallymeanit
      prompt: Are you sure you want to shrink the cluster?
      default: 'no'
      private: no
  tasks:
    - name: exit playbook, if no rgw was given
      when: rgw_to_kill is not defined or rgw_to_kill | length == 0
      fail:
        msg: >
          rgw_to_kill must be declared.
          Exiting shrink-cluster playbook, no RGW was removed. On the command
          line when invoking the playbook, you can use
          "-e rgw_to_kill=ceph.rgw0 argument".  You can only remove a single
          RGW each time the playbook runs.          

    - name: exit playbook, if user did not mean to shrink cluster
      when: ireallymeanit != 'yes'
      fail:
        msg: >
          Exiting shrink-mon playbook, no monitor was removed. To shrink the
          cluster, either say 'yes' on the prompt or use
          '-e ireallymeanit=yes' on the command line when  invoking the playbook          

- name: gather facts and mons and rgws
  hosts:
    - "{{ mon_group_name | default('mons') }}[0]"
    - "{{ rgw_group_name | default('rgws') }}"
  become: true
  gather_facts: false
  tasks:
    - name: gather facts
      setup:
        gather_subset:
          - 'all'
          - '!facter'
          - '!ohai'

- hosts: mons[0]
  become: true
  gather_facts: false
  pre_tasks:
    - import_role:
        name: ceph-defaults

    - import_role:
        name: ceph-facts
        tasks_from: container_binary

    - name: set_fact container_exec_cmd for mon0
      set_fact:
        container_exec_cmd: "{{ container_binary }} exec ceph-mon-{{ ansible_facts['hostname'] }}"
      when: containerized_deployment | bool

    - name: exit playbook, if can not connect to the cluster
      command: "{{ container_exec_cmd | default('') }} timeout 5 ceph --cluster {{ cluster }} health"
      register: ceph_health
      changed_when: false
      until: ceph_health is succeeded
      retries: 5
      delay: 2

    - name: get rgw instances
      command: "{{ container_exec_cmd | default('') }} ceph --cluster {{ cluster }} service dump -f json"
      register: rgw_instances
      changed_when: false


    - name: exit playbook, if the rgw_to_kill doesn't exist
      when: rgw_to_kill not in (rgw_instances.stdout | from_json).services.rgw.daemons.keys() | list
      fail:
        msg: >
          It seems that the rgw instance given is not part of the ceph cluster. Please
          make sure it is.
          The rgw instance format is $(hostname}.rgw$(instance number).          
  tasks:
    - name: get rgw host running the rgw instance to kill
      set_fact:
        rgw_host: '{{ item }}'
      with_items: '{{ groups[rgw_group_name] }}'
      when: hostvars[item]['ansible_facts']['hostname'] == rgw_to_kill.split('.')[0]

    - name: stop rgw service
      service:
        name: ceph-radosgw@rgw.{{ rgw_to_kill }}
        state: stopped
        enabled: no
      delegate_to: "{{ rgw_host }}"
      failed_when: false

    - name: ensure that the rgw is stopped
      command: "systemctl is-active ceph-radosgw@rgw.{{ rgw_to_kill }}"  # noqa 303
      register: rgw_to_kill_status
      failed_when: rgw_to_kill_status.rc == 0
      changed_when: false
      delegate_to: "{{ rgw_host }}"
      retries: 5
      delay: 2

    - name: exit if rgw_to_kill is reported in ceph status
      command: "{{ container_exec_cmd | default('') }} ceph --cluster {{ cluster }} service dump -f json"
      register: ceph_status
      changed_when: false
      failed_when:
        - (ceph_status.stdout | from_json).services.rgw is defined
        - rgw_to_kill in (ceph_status.stdout | from_json).services.rgw.daemons.keys() | list
      until:
        - (ceph_status.stdout | from_json).services.rgw is defined
        - rgw_to_kill not in (ceph_status.stdout | from_json).services.rgw.daemons.keys() | list
      retries: 3
      delay: 3

    - name: purge directories related to rgw
      file:
        path: /var/lib/ceph/radosgw/{{ cluster }}-rgw.{{ rgw_to_kill }}
        state: absent
      delegate_to: "{{ rgw_host }}"
  post_tasks:
    - name: show ceph health
      command: "{{ container_exec_cmd | default('') }} ceph --cluster {{ cluster }} -s"
      changed_when: false