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

125 lines
4.6 KiB
YAML
Raw Normal View History

---
# This playbook removes the Ceph RBD mirror from your cluster on the given
# node.
#
# Use it like this:
# ansible-playbook shrink-rbdmirror.yml -e rbdmirror_to_kill=ceph-rbdmirror01
# 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-rbdmirror.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:
- mons
- rbdmirrors
become: true
tasks:
- debug:
msg: gather facts on MONs and RBD mirrors
- name: confirm whether user really meant to remove rbd mirror from the ceph
cluster
hosts: mons[0]
become: true
vars_prompt:
- name: ireallymeanit
prompt: Are you sure you want to shrink the cluster?
default: 'no'
private: no
pre_tasks:
- import_role:
name: ceph-defaults
- import_role:
name: ceph-facts
tasks_from: container_binary
- name: exit playbook, if no rbdmirror was given
fail:
msg: "rbdmirror_to_kill must be declared
Exiting shrink-cluster playbook, no RBD mirror was removed.
On the command line when invoking the playbook, you can use
-e rbdmirror_to_kill=rbd-mirror01 argument. You can only remove a
single rbd mirror each time the playbook runs."
when: rbdmirror_to_kill is not defined
- name: exit playbook, if the rbdmirror is not part of the inventory
fail:
msg: >
It seems that the host given is not part of your inventory,
please make sure it is.
when: rbdmirror_to_kill not in groups[rbdmirror_group_name]
- name: exit playbook, if user did not mean to shrink cluster
fail:
msg: "Exiting shrink-rbdmirror playbook, no rbd-mirror was 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: set_fact container_exec_cmd for mon0
when: containerized_deployment | bool
set_fact:
container_exec_cmd: "{{ container_binary }} exec ceph-mon-{{ ansible_facts['hostname'] }}"
- name: exit playbook, if can not connect to the cluster
command: "{{ container_exec_cmd | default('') }} timeout 5 ceph --cluster {{ cluster }} service dump -f json"
register: ceph_health
changed_when: false
until: ceph_health is succeeded
retries: 5
delay: 2
- name: set_fact rbdmirror_to_kill_hostname
set_fact:
rbdmirror_to_kill_hostname: "{{ hostvars[rbdmirror_to_kill]['ansible_facts']['hostname'] }}"
- name: set_fact rbdmirror_gids
set_fact:
rbdmirror_gids: "{{ rbdmirror_gids | default([]) + [ item ] }}"
with_items: "{{ (ceph_health.stdout | from_json)['services']['rbd-mirror']['daemons'].keys() | list }}"
when: item != 'summary'
- name: set_fact rbdmirror_to_kill_gid
set_fact:
rbdmirror_to_kill_gid: "{{ (ceph_health.stdout | from_json)['services']['rbd-mirror']['daemons'][item]['gid'] }}"
with_items: "{{ rbdmirror_gids }}"
when: (ceph_health.stdout | from_json)['services']['rbd-mirror']['daemons'][item]['metadata']['id'] == rbdmirror_to_kill_hostname
tasks:
- name: stop rbdmirror service
service:
name: ceph-rbd-mirror@rbd-mirror.{{ rbdmirror_to_kill_hostname }}
state: stopped
enabled: no
delegate_to: "{{ rbdmirror_to_kill }}"
failed_when: false
- name: purge related directories
file:
path: /var/lib/ceph/bootstrap-rbd-mirror/{{ cluster }}-{{ rbdmirror_to_kill_hostname }}
state: absent
delegate_to: "{{ rbdmirror_to_kill }}"
post_tasks:
- name: get servicemap details
command: "{{ container_exec_cmd | default('') }} timeout 5 ceph --cluster {{ cluster }} service dump -f json"
register: ceph_health
failed_when:
- "'rbd-mirror' in (ceph_health.stdout | from_json)['services'].keys() | list"
- rbdmirror_to_kill_gid in (ceph_health.stdout | from_json)['services']['rbd-mirror']['daemons'].keys() | list
until:
- "'rbd-mirror' in (ceph_health.stdout | from_json)['services'].keys() | list"
- rbdmirror_to_kill_gid not in (ceph_health.stdout | from_json)['services']['rbd-mirror']['daemons'].keys() | list
when: rbdmirror_to_kill_gid is defined
retries: 12
delay: 10
- name: show ceph health
command: "{{ container_exec_cmd | default('') }} ceph --cluster {{ cluster }} -s"
changed_when: false