From 86f3eeb717c7daac8c6330fdaa7f8a3c83f94b0d Mon Sep 17 00:00:00 2001
From: Guillaume Abrioux <gabrioux@redhat.com>
Date: Thu, 9 Jan 2020 16:46:34 +0100
Subject: [PATCH] mon: support replacing a mon

We must pick up a mon which actually exists in ceph-facts in order to
detect if a cluster is running. Otherwise, it will state no cluster is
already running which will end up deploying a new monitor isolated in a
new quorum.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1622688

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
---
 roles/ceph-facts/tasks/facts.yml         | 61 +++++++++++++++++++++++-
 roles/ceph-mon/tasks/deploy_monitors.yml | 10 ++--
 2 files changed, 64 insertions(+), 7 deletions(-)

diff --git a/roles/ceph-facts/tasks/facts.yml b/roles/ceph-facts/tasks/facts.yml
index ff5375458..6a9418578 100644
--- a/roles/ceph-facts/tasks/facts.yml
+++ b/roles/ceph-facts/tasks/facts.yml
@@ -45,17 +45,74 @@
     - containerized_deployment | bool
     - groups.get(mon_group_name, []) | length > 0
 
+- name: find a running mon container
+  command: "{{ container_binary }} ps -q --filter name=ceph-mon-{{ hostvars[item]['ansible_hostname'] }}"
+  register: find_running_mon_container
+  failed_when: false
+  run_once: true
+  delegate_to: "{{ item }}"
+  with_items: "{{ groups.get(mon_group_name, []) }}"
+  when: containerized_deployment | bool
+
+- name: check for a ceph mon socket
+  shell: stat --printf=%n {{ rbd_client_admin_socket_path }}/{{ cluster }}-mon*.asok
+  changed_when: false
+  failed_when: false
+  check_mode: no
+  register: mon_socket_stat
+  run_once: true
+  delegate_to: "{{ item }}"
+  with_items: "{{ groups.get(mon_group_name, []) }}"
+  when: not containerized_deployment | bool
+
+- name: check if the ceph mon socket is in-use
+  command: grep -q {{ item.stdout }} /proc/net/unix
+  changed_when: false
+  failed_when: false
+  check_mode: no
+  register: mon_socket
+  run_once: true
+  with_items: "{{ mon_socket_stat.results }}"
+  when:
+    - not containerized_deployment | bool
+    - item.rc == 0
+
+- name: set_fact running_mon - non_container
+  set_fact:
+    running_mon: "{{ hostvars[item.item.item]['inventory_hostname'] }}"
+  with_items: "{{ mon_socket.results }}"
+  run_once: true
+  when:
+    - not containerized_deployment | bool
+    - item.rc is defined
+    - item.rc == 1
+
+- name: set_fact running_mon - container
+  set_fact:
+    running_mon: "{{ item.item }}"
+  run_once: true
+  with_items: "{{ find_running_mon_container.results }}"
+  when:
+    - containerized_deployment | bool
+    - item.stdout_lines | default([]) | length > 0
+
+- name: set_fact _container_exec_cmd
+  set_fact:
+    _container_exec_cmd: "{{ container_binary }} exec ceph-mon-{{ hostvars[groups[mon_group_name][0]]['ansible_hostname'] if running_mon is undefined else running_mon }}"
+  when:
+    - containerized_deployment | bool
+
 # this task shouldn't run in a rolling_update situation
 # because it blindly picks a mon, which may be down because
 # of the rolling update
 - name: is ceph running already?
-  command: "{{ timeout_command }} {{ container_exec_cmd }} ceph --cluster {{ cluster }} -s -f json"
+  command: "{{ timeout_command }} {{ _container_exec_cmd | default('') }} ceph --cluster {{ cluster }} -s -f json"
   changed_when: false
   failed_when: false
   check_mode: no
   register: ceph_current_status
   run_once: true
-  delegate_to: "{{ groups[mon_group_name][0] }}"
+  delegate_to: "{{ groups[mon_group_name][0] if running_mon is undefined else running_mon }}"
   when:
     - not rolling_update | bool
     - groups.get(mon_group_name, []) | length > 0
diff --git a/roles/ceph-mon/tasks/deploy_monitors.yml b/roles/ceph-mon/tasks/deploy_monitors.yml
index 0dfeacec4..e9cce2c50 100644
--- a/roles/ceph-mon/tasks/deploy_monitors.yml
+++ b/roles/ceph-mon/tasks/deploy_monitors.yml
@@ -1,23 +1,23 @@
 ---
 - name: check if monitor initial keyring already exists
   command: >
-    {{ hostvars[groups[mon_group_name][0]]['container_exec_cmd'] | default('') }} ceph --cluster {{ cluster }} --name mon. -k
-    /var/lib/ceph/mon/{{ cluster }}-{{ hostvars[groups[mon_group_name][0]]['ansible_hostname'] }}/keyring
+    {{ _container_exec_cmd | default('') }} ceph --cluster {{ cluster }} --name mon. -k
+    /var/lib/ceph/mon/{{ cluster }}-{{ hostvars[groups[mon_group_name][0] if running_mon is undefined else running_mon]['ansible_hostname'] }}/keyring
     auth get-key mon.
   register: initial_mon_key
   run_once: True
-  delegate_to: "{{ groups.get(mon_group_name, [])[0] }}"
+  delegate_to: "{{ groups[mon_group_name][0] if running_mon is undefined else running_mon }}"
   when: ceph_current_status.fsid is defined
 
 - name: generate monitor initial keyring
   command: >
-    {{ discovered_interpreter_python }} -c "import os ; import struct ;
+    {{ hostvars[groups[mon_group_name][0] if running_mon is undefined else running_mon]['discovered_interpreter_python'] }} -c "import os ; import struct ;
     import time; import base64 ; key = os.urandom(16) ;
     header = struct.pack('<hiih',1,int(time.time()),0,len(key)) ;
     print(base64.b64encode(header + key).decode())"
   register: monitor_keyring
   run_once: True
-  delegate_to: "{{ groups.get(mon_group_name, [])[0] }}"
+  delegate_to: "{{ groups[mon_group_name][0] if running_mon is undefined else running_mon }}"
   when:
     - initial_mon_key.skipped is defined
     - ceph_current_status.fsid is undefined
-- 
2.39.5