ceph-ansible/infrastructure-playbooks/untested-by-ci/replace-osd.yml

---
# This playbook replaces Ceph OSDs.
# It can replace any number of OSD(s) from the cluster and ALL THEIR DATA
#
# When disks fail, or if an admnistrator wants to reprovision OSDs with a new backend,
# for instance, for switching from FileStore to BlueStore, OSDs need to be replaced.
# Unlike Removing the OSD, replaced OSD’s id and CRUSH map entry need to be keep intact after the OSD is destroyed for replacement.
#
# Use it like this:
# ansible-playbook replace-osd.yml -e osd_to_replace=0,2,6
#     Prompts for confirmation to replace, defaults to no and
#     doesn't replace the osd(s). yes replaces the osd(s).
#
# ansible-playbook -e ireallymeanit=yes|no replace-osd.yml
#     Overrides the prompt using -e option. Can be used in
#     automation scripts to avoid interactive prompt.

- name: Gather facts and check the init system

  hosts:
    - "{{ mon_group_name|default('mons') }}"
    - "{{ osd_group_name|default('osds') }}"

  become: true
  tasks:
    - ansible.builtin.debug: msg="gather facts on all Ceph hosts for following reference"

- name: Confirm whether user really meant to replace osd(s)
  hosts: localhost
  become: true
  vars_prompt:
    - name: ireallymeanit  # noqa: name[casing]
      prompt: Are you sure you want to replace the osd(s)?
      default: 'no'
      private: false
  vars:
    mon_group_name: mons
    osd_group_name: osds
  pre_tasks:
    - name: Exit playbook, if user did not mean to replace the osd(s)
      ansible.builtin.fail:
        msg: "Exiting replace-osd playbook, no osd(s) was/were replaced..
           To replace the osd(s), either say 'yes' on the prompt or
           or use `-e ireallymeanit=yes` on the command line when
           invoking the playbook"
      when: ireallymeanit != 'yes'

    - name: Exit playbook, if no osd(s) was/were given
      ansible.builtin.fail:
        msg: "osd_to_replace must be declared
          Exiting replace-osd playbook, no OSD(s) was/were replaced.
           On the command line when invoking the playbook, you can use
           -e osd_to_replace=0,1,2,3 argument."
      when: osd_to_replace is not defined

  tasks:
    - ansible.builtin.import_role:
        name: ceph-defaults

  post_tasks:
    - name: Set_fact container_exec_cmd build docker exec command (containerized)
      ansible.builtin.set_fact:
        container_exec_cmd: "docker exec ceph-mon-{{ hostvars[groups[mon_group_name][0]]['ansible_facts']['hostname'] }}"
      when: containerized_deployment | bool

    - name: Exit playbook, if can not connect to the cluster
      ansible.builtin.command: "{{ container_exec_cmd | default('') }} timeout 5 ceph --cluster {{ cluster }} health"
      register: ceph_health
      until: ceph_health.stdout.find("HEALTH") > -1
      delegate_to: "{{ groups[mon_group_name][0] }}"
      retries: 5
      delay: 2

    - name: Find the host(s) where the osd(s) is/are running on
      ansible.builtin.command: "{{ container_exec_cmd | default('') }} ceph --cluster {{ cluster }} osd find {{ item }}"
      with_items: "{{ osd_to_replace.split(',') }}"
      delegate_to: "{{ groups[mon_group_name][0] }}"
      register: find_osd_hosts

    - name: Set_fact osd_hosts
      ansible.builtin.set_fact:
        osd_hosts: "{{ osd_hosts | default([]) + [ (item.stdout | from_json).crush_location.host ] }}"
      with_items: "{{ find_osd_hosts.results }}"

    - name: Check if ceph admin key exists on the osd nodes
      ansible.builtin.stat:
        path: "/etc/ceph/{{ cluster }}.client.admin.keyring"
      register: ceph_admin_key
      with_items: "{{ osd_hosts }}"
      delegate_to: "{{ item }}"
      failed_when: false
      when: not containerized_deployment | bool

    - name: Fail when admin key is not present
      ansible.builtin.fail:
        msg: "The Ceph admin key is not present on the OSD node, please add it and remove it after the playbook is done."
      with_items: "{{ ceph_admin_key.results }}"
      when:
        - not containerized_deployment | bool
        - item.stat.exists == false

    # NOTE(leseb): using '>' is the only way I could have the command working
    - name: Find osd device based on the id
      ansible.builtin.shell: >
        docker run --privileged=true -v /dev:/dev --entrypoint /usr/sbin/ceph-disk
        {{ ceph_docker_registry}}/{{ ceph_docker_image }}:{{ ceph_docker_image_tag }}
        list | awk  -v pattern=osd.{{ item.1 }} '$0 ~ pattern {print $1}'        
      with_together:
        - "{{ osd_hosts }}"
        - "{{ osd_to_replace.split(',') }}"
      register: osd_to_replace_disks
      delegate_to: "{{ item.0 }}"
      when: containerized_deployment | bool

    - name: Zapping osd(s) - container
      ansible.builtin.shell: >
        docker run --privileged=true -v /dev:/dev --entrypoint /usr/sbin/ceph-disk
        {{ ceph_docker_registry}}/{{ ceph_docker_image }}:{{ ceph_docker_image_tag }}
        zap {{ item.1 }}        
      run_once: true
      with_together:
        - "{{ osd_hosts }}"
        - "{{ osd_to_replace_disks.results }}"
      delegate_to: "{{ item.0 }}"
      when: containerized_deployment | bool

    - name: Zapping osd(s) - non container
      ansible.builtin.command: ceph-disk zap --cluster {{ cluster }} {{ item.1 }}
      run_once: true
      with_together:
        - "{{ osd_hosts }}"
        - "{{ osd_to_replace_disks.results }}"
      delegate_to: "{{ item.0 }}"
      when: not containerized_deployment | bool

    - name: Destroying osd(s)
      ansible.builtin.command: ceph-disk destroy --cluster {{ cluster }} --destroy-by-id {{ item.1 }} --zap
      run_once: true
      with_together:
        - "{{ osd_hosts }}"
        - "{{ osd_to_replace.split(',') }}"
      delegate_to: "{{ item.0 }}"
      when: not containerized_deployment | bool

    - name: Replace osd(s) - prepare - non container
      ansible.builtin.command: ceph-disk prepare {{ item.1 }}  --osd-id {{ item.2 }} --osd-uuid $(uuidgen)
      run_once: true
      delegate_to: "{{ item.0 }}"
      with_together:
        - "{{ osd_hosts }}"
        - "{{ osd_to_replace_disks.results }}"
        - "{{ osd_to_replace.split(',') }}"

    - name: Replace osd(s) - prepare - container
      ansible.builtin.shell: >
        docker run --privileged=true -v /dev:/dev --entrypoint /usr/sbin/ceph-disk
        {{ ceph_docker_registry}}/{{ ceph_docker_image }}:{{ ceph_docker_image_tag }}
        prepare {{ item.1 }}        
      run_once: true
      delegate_to: "{{ item.0 }}"
      with_together:
        - "{{ osd_hosts }}"
        - "{{ osd_to_replace_disks.results }}"

    - name: Replace osd(s) - activate - non container
      ansible.builtin.command: ceph-disk activate {{ item.1 }}1
      run_once: true
      delegate_to: "{{ item.0 }}"
      with_together:
        - "{{ osd_hosts }}"
        - "{{ osd_to_replace_disks.results }}"

    - name: Replace osd(s) - activate - container
      ansible.builtin.shell: >
        docker run --privileged=true -v /dev:/dev --entrypoint /usr/sbin/ceph-disk
        {{ ceph_docker_registry}}/{{ ceph_docker_image }}:{{ ceph_docker_image_tag }}
        activate {{ item.1 }}1        
      run_once: true
      delegate_to: "{{ item.0 }}"
      with_together:
        - "{{ osd_hosts }}"
        - "{{ osd_to_replace_disks.results }}"

    - name: Show ceph health
      ansible.builtin.command: "{{ container_exec_cmd | default('') }} ceph --cluster {{ cluster }} -s"
      delegate_to: "{{ groups[mon_group_name][0] }}"

    - name: Show ceph osd tree
      ansible.builtin.command: "{{ container_exec_cmd | default('') }} ceph --cluster {{ cluster }} osd tree"
      delegate_to: "{{ groups[mon_group_name][0] }}"