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

128 lines
4.4 KiB
YAML

---
# This playbook shrinks Ceph OSDs that have been created with ceph-volume.
# It can remove any number of OSD(s) from the cluster and ALL THEIR DATA
#
# Use it like this:
# ansible-playbook shrink-osd.yml -e osd_to_kill=0,2,6
# 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-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:
- debug: msg="gather facts on all Ceph hosts for following reference"
- name: confirm whether user really meant to remove osd(s) from the cluster
hosts: "{{ groups[mon_group_name][0] }}"
become: true
vars_prompt:
- name: ireallymeanit
prompt: Are you sure you want to shrink the cluster?
default: 'no'
private: no
vars:
mon_group_name: mons
osd_group_name: osds
pre_tasks:
- name: exit playbook, if user did not mean to shrink cluster
fail:
msg: "Exiting shrink-osd playbook, no osd(s) was/were removed..
To shrink the cluster, 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
fail:
msg: "osd_to_kill must be declared
Exiting shrink-osd playbook, no OSD(s) was/were removed.
On the command line when invoking the playbook, you can use
-e osd_to_kill=0,1,2,3 argument."
when: osd_to_kill is not defined
tasks:
- import_role:
name: ceph-defaults
- import_role:
name: ceph-facts
post_tasks:
- name: set_fact container_exec_cmd build docker exec command (containerized)
set_fact:
container_exec_cmd: "{{ container_binary }} exec ceph-mon-{{ hostvars[groups[mon_group_name][0]]['ansible_hostname'] }}"
when: containerized_deployment | bool
- name: exit playbook, if can not connect to the cluster
command: "{{ container_exec_cmd }} timeout 5 ceph --cluster {{ cluster }} health"
register: ceph_health
until: ceph_health.stdout.find("HEALTH") > -1
retries: 5
delay: 2
- name: find the host(s) where the osd(s) is/are running on
command: "{{ container_exec_cmd }} ceph --cluster {{ cluster }} osd find {{ item }}"
with_items: "{{ osd_to_kill.split(',') }}"
register: find_osd_hosts
- name: set_fact osd_hosts
set_fact:
osd_hosts: "{{ osd_hosts | default([]) + [ [ (item.stdout | from_json).crush_location.host, (item.stdout | from_json).osd_fsid ] ] }}"
with_items: "{{ find_osd_hosts.results }}"
- name: set_fact _osd_hosts
set_fact:
_osd_hosts: "{{ _osd_hosts | default([]) + [ [ item.0, item.2 ] ] }}"
with_nested:
- "{{ groups.get(osd_group_name) }}"
- "{{ osd_hosts }}"
when: hostvars[item.0]['ansible_hostname'] == item.1
- name: mark osd(s) out of the cluster
command: "{{ container_exec_cmd }} ceph --cluster {{ cluster }} osd out {{ osd_to_kill.replace(',', ' ') }}"
run_once: true
- name: stop osd(s) service
service:
name: ceph-osd@{{ item.0 }}
state: stopped
enabled: no
loop: "{{ osd_to_kill.split(',')|zip(_osd_hosts)|list }}"
delegate_to: "{{ item.1.0 }}"
- name: zap osd devices
ceph_volume:
action: "zap"
osd_fsid: "{{ item.1 }}"
environment:
CEPH_VOLUME_DEBUG: 1
CEPH_CONTAINER_IMAGE: "{{ ceph_docker_registry + '/' + ceph_docker_image + ':' + ceph_docker_image_tag if containerized_deployment else None }}"
CEPH_CONTAINER_BINARY: "{{ container_binary }}"
delegate_to: "{{ item.0 }}"
loop: "{{ _osd_hosts }}"
- name: purge osd(s) from the cluster
command: "{{ container_exec_cmd }} ceph --cluster {{ cluster }} osd purge {{ item }} --yes-i-really-mean-it"
run_once: true
with_items: "{{ osd_to_kill.split(',') }}"
- name: show ceph health
command: "{{ container_exec_cmd }} ceph --cluster {{ cluster }} -s"
- name: show ceph osd tree
command: "{{ container_exec_cmd }} ceph --cluster {{ cluster }} osd tree"