ceph-ansible/infrastructure-playbooks/untested-by-ci/cluster-os-migration.yml

---
# This playbook was meant to upgrade a node from Ubuntu to RHEL.
# We are performing a set of actions prior to reboot the node.
# The node reboots via PXE and gets its new operating system.
# This playbook only works for monitors and OSDs.
# Note that some of the checks are ugly:
#   ie: the when migration_completed.stat.exists
# can be improved with includes, however I wanted to keep a single file...
#

- hosts: mons
  serial: 1
  sudo: True

  vars:
    backup_dir: /tmp/

  tasks:

    - name: Check if the node has be migrated already
      stat: >
        path=/var/lib/ceph/mon/ceph-{{ ansible_facts['hostname'] }}/migration_completed
      register: migration_completed
      failed_when: false

    - name: Check for failed run
      stat: >
        path=/var/lib/ceph/{{ ansible_facts['hostname'] }}.tar
      register: mon_archive_leftover

    - fail: msg="Looks like an archive is already there, please remove it!"
      when: migration_completed.stat.exists == False and mon_archive_leftover.stat.exists == True

    - name: Compress the store as much as possible
      command: ceph tell mon.{{ ansible_facts['hostname'] }} compact
      when: migration_completed.stat.exists == False

    - name: Check if sysvinit
      stat: >
        path=/var/lib/ceph/mon/ceph-{{ ansible_facts['hostname'] }}/sysvinit
      register: monsysvinit
      changed_when: False

    - name: Check if upstart
      stat: >
        path=/var/lib/ceph/mon/ceph-{{ ansible_facts['hostname'] }}/upstart
      register: monupstart
      changed_when: False

    - name: Check if init does what it is supposed to do (Sysvinit)
      shell: >
        ps faux|grep -sq [c]eph-mon && service ceph status mon >> /dev/null
      register: ceph_status_sysvinit
      changed_when: False

    # can't complete the condition since the previous taks never ran...
    - fail: msg="Something is terribly wrong here, sysvinit is configured, the service is started BUT the init script does not return 0, GO FIX YOUR SETUP!"
      when: ceph_status_sysvinit.rc != 0 and migration_completed.stat.exists == False and monsysvinit.stat.exists == True

    - name: Check if init does what it is supposed to do (upstart)
      shell: >
        ps faux|grep -sq [c]eph-mon && status ceph-mon-all >> /dev/null
      register: ceph_status_upstart
      changed_when: False

    - fail: msg="Something is terribly wrong here, upstart is configured, the service is started BUT the init script does not return 0, GO FIX YOUR SETUP!"
      when: ceph_status_upstart.rc != 0 and migration_completed.stat.exists == False and monupstart.stat.exists == True

    - name: Restart the Monitor after compaction (Upstart)
      service: >
        name=ceph-mon
        state=restarted
        args=id={{ ansible_facts['hostname'] }}
      when: monupstart.stat.exists == True and migration_completed.stat.exists == False

    - name: Restart the Monitor after compaction (Sysvinit)
      service: >
        name=ceph
        state=restarted
        args=mon
      when: monsysvinit.stat.exists == True and migration_completed.stat.exists == False

    - name: Wait for the monitor to be up again
      local_action:
        module: wait_for
        host: "{{ ansible_ssh_host | default(inventory_hostname) }}"
        port: 6789
        timeout: 10
      when: migration_completed.stat.exists == False

    - name: Stop the monitor (Upstart)
      service: >
        name=ceph-mon
        state=stopped
        args=id={{ ansible_facts['hostname'] }}
      when: monupstart.stat.exists == True and migration_completed.stat.exists == False

    - name: Stop the monitor (Sysvinit)
      service: >
        name=ceph
        state=stopped
        args=mon
      when: monsysvinit.stat.exists == True and migration_completed.stat.exists == False

    - name: Wait for the monitor to be down
      local_action:
        module: wait_for
        host: "{{ ansible_ssh_host | default(inventory_hostname) }}"
        port: 6789
        timeout: 10
        state: stopped
      when: migration_completed.stat.exists == False

    - name: Create a backup directory
      file: >
        path={{ backup_dir }}/monitors-backups
        state=directory
        owner=root
        group=root
        mode=0644
      delegate_to: "{{ item }}"
      with_items: "{{ groups.backup[0] }}"
      when: migration_completed.stat.exists == False

    # NOTE (leseb): should we convert upstart to sysvinit here already?
    - name: Archive monitor stores
      shell: >
        tar -cpvzf - --one-file-system . /etc/ceph/* | cat > {{ ansible_facts['hostname'] }}.tar
        chdir=/var/lib/ceph/
        creates={{ ansible_facts['hostname'] }}.tar
      when: migration_completed.stat.exists == False

    - name: Scp the Monitor store
      fetch: >
        src=/var/lib/ceph/{{ ansible_facts['hostname'] }}.tar
        dest={{ backup_dir }}/monitors-backups/{{ ansible_facts['hostname'] }}.tar
        flat=yes
      when: migration_completed.stat.exists == False

    - name: Reboot the server
      command: reboot
      when: migration_completed.stat.exists == False

    - name: Wait for the server to come up
      local_action:
        module: wait_for
        port: 22
        delay: 10
        timeout: 3600
      when: migration_completed.stat.exists == False

    - name: Wait a bit more to be sure that the server is ready
      pause: seconds=20
      when: migration_completed.stat.exists == False

    - name: Check if sysvinit
      stat: >
        path=/var/lib/ceph/mon/ceph-{{ ansible_facts['hostname'] }}/sysvinit
      register: monsysvinit
      changed_when: False

    - name: Check if upstart
      stat: >
        path=/var/lib/ceph/mon/ceph-{{ ansible_facts['hostname'] }}/upstart
      register: monupstart
      changed_when: False

    - name: Make sure the monitor is stopped (Upstart)
      service: >
        name=ceph-mon
        state=stopped
        args=id={{ ansible_facts['hostname'] }}
      when: monupstart.stat.exists == True and migration_completed.stat.exists == False

    - name: Make sure the monitor is stopped (Sysvinit)
      service: >
        name=ceph
        state=stopped
        args=mon
      when: monsysvinit.stat.exists == True and migration_completed.stat.exists == False

    # NOTE (leseb): 'creates' was added in Ansible 1.6
    - name: Copy and unarchive the monitor store
      unarchive: >
        src={{ backup_dir }}/monitors-backups/{{ ansible_facts['hostname'] }}.tar
        dest=/var/lib/ceph/
        copy=yes
        mode=0600
        creates=etc/ceph/ceph.conf
      when: migration_completed.stat.exists == False

    - name: Copy keys and configs
      shell: >
        cp etc/ceph/* /etc/ceph/
        chdir=/var/lib/ceph/
      when: migration_completed.stat.exists == False

    - name: Configure RHEL7 for sysvinit
      shell: find -L /var/lib/ceph/mon/ -mindepth 1 -maxdepth 1 -regextype posix-egrep -regex '.*/[A-Za-z0-9]+-[A-Za-z0-9._-]+' -exec touch {}/sysvinit \; -exec rm {}/upstart \;
      when: migration_completed.stat.exists == False

    # NOTE (leseb): at this point the upstart and sysvinit checks are not necessary
    # so we directly call sysvinit
    - name: Start the monitor
      service: >
        name=ceph
        state=started
        args=mon
      when: migration_completed.stat.exists == False

    - name: Wait for the Monitor to be up again
      local_action:
        module: wait_for
        host: "{{ ansible_ssh_host | default(inventory_hostname) }}"
        port: 6789
        timeout: 10
      when: migration_completed.stat.exists == False

    - name: Waiting for the monitor to join the quorum...
      shell: >
        ceph -s | grep monmap | sed 's/.*quorum//' | egrep -q {{ ansible_facts['hostname'] }}
      register: result
      until: result.rc == 0
      retries: 5
      delay: 10
      delegate_to: "{{ item }}"
      with_items: "{{ groups.backup[0] }}"
      when: migration_completed.stat.exists == False

    - name: Done moving to the next monitor
      file: >
        path=/var/lib/ceph/mon/ceph-{{ ansible_facts['hostname'] }}/migration_completed
        state=touch
        owner=root
        group=root
        mode=0600
      when: migration_completed.stat.exists == False

- hosts: osds
  serial: 1
  sudo: True

  vars:
    backup_dir: /tmp/

  tasks:
    - name: Check if the node has be migrated already
      stat: >
        path=/var/lib/ceph/migration_completed
      register: migration_completed
      failed_when: false

    - name: Check for failed run
      stat: >
        path=/var/lib/ceph/{{ ansible_facts['hostname'] }}.tar
      register: osd_archive_leftover

    - fail: msg="Looks like an archive is already there, please remove it!"
      when: migration_completed.stat.exists == False and osd_archive_leftover.stat.exists == True

    - name: Check if init does what it is supposed to do (Sysvinit)
      shell: >
        ps faux|grep -sq [c]eph-osd && service ceph status osd >> /dev/null
      register: ceph_status_sysvinit
      changed_when: False

    # can't complete the condition since the previous taks never ran...
    - fail: msg="Something is terribly wrong here, sysvinit is configured, the services are started BUT the init script does not return 0, GO FIX YOUR SETUP!"
      when: ceph_status_sysvinit.rc != 0 and migration_completed.stat.exists == False and monsysvinit.stat.exists == True

    - name: Check if init does what it is supposed to do (upstart)
      shell: >
        ps faux|grep -sq [c]eph-osd && initctl list|egrep -sq "ceph-osd \(ceph/.\) start/running, process [0-9][0-9][0-9][0-9]"
      register: ceph_status_upstart
      changed_when: False

    - fail: msg="Something is terribly wrong here, upstart is configured, the services are started BUT the init script does not return 0, GO FIX YOUR SETUP!"
      when: ceph_status_upstart.rc != 0 and migration_completed.stat.exists == False and monupstart.stat.exists == True

    - name: Set the noout flag
      command: ceph osd set noout
      delegate_to: "{{ item }}"
      with_items: "{{ groups[mon_group_name][0] }}"
      when: migration_completed.stat.exists == False

    - name: Check if sysvinit
      shell: stat /var/lib/ceph/osd/ceph-*/sysvinit
      register: osdsysvinit
      failed_when: false
      changed_when: False

    - name: Check if upstart
      shell: stat /var/lib/ceph/osd/ceph-*/upstart
      register: osdupstart
      failed_when: false
      changed_when: False

    - name: Archive ceph configs
      shell: >
        tar -cpvzf - --one-file-system . /etc/ceph/ceph.conf | cat > {{ ansible_facts['hostname'] }}.tar
        chdir=/var/lib/ceph/
        creates={{ ansible_facts['hostname'] }}.tar
      when: migration_completed.stat.exists == False

    - name: Create backup directory
      file: >
        path={{ backup_dir }}/osds-backups
        state=directory
        owner=root
        group=root
        mode=0644
      delegate_to: "{{ item }}"
      with_items: "{{ groups.backup[0] }}"
      when: migration_completed.stat.exists == False

    - name: Scp OSDs dirs and configs
      fetch: >
        src=/var/lib/ceph/{{ ansible_facts['hostname'] }}.tar
        dest={{ backup_dir }}/osds-backups/
        flat=yes
      when: migration_completed.stat.exists == False

    - name: Collect OSD ports
      shell: netstat -tlpn | awk -F ":" '/ceph-osd/ { sub (" .*", "", $2); print $2 }' | uniq
      register: osd_ports
      when: migration_completed.stat.exists == False

    - name: Gracefully stop the OSDs (Upstart)
      service: >
        name=ceph-osd-all
        state=stopped
      when: osdupstart.rc == 0 and migration_completed.stat.exists == False

    - name: Gracefully stop the OSDs (Sysvinit)
      service: >
        name=ceph
        state=stopped
        args=mon
      when: osdsysvinit.rc == 0 and migration_completed.stat.exists == False

    - name: Wait for the OSDs to be down
      local_action:
        module: wait_for
        host: "{{ ansible_ssh_host | default(inventory_hostname) }}"
        port: {{ item }}
        timeout: 10
        state: stopped
      with_items: "{{ osd_ports.stdout_lines }}"
      when: migration_completed.stat.exists == False

    - name: Configure RHEL with sysvinit
      shell: find -L /var/lib/ceph/osd/ -mindepth 1 -maxdepth 1 -regextype posix-egrep -regex '.*/[A-Za-z0-9]+-[A-Za-z0-9._-]+' -exec touch {}/sysvinit \; -exec rm {}/upstart \;
      when: migration_completed.stat.exists == False

    - name: Reboot the server
      command: reboot
      when: migration_completed.stat.exists == False

    - name: Wait for the server to come up
      local_action:
        module: wait_for
        port: 22
        delay: 10
        timeout: 3600
      when: migration_completed.stat.exists == False

    - name: Wait a bit to be sure that the server is ready for scp
      pause: seconds=20
      when: migration_completed.stat.exists == False

    # NOTE (leseb): 'creates' was added in Ansible 1.6
    - name: Copy and unarchive the OSD configs
      unarchive: >
        src={{ backup_dir }}/osds-backups/{{ ansible_facts['hostname'] }}.tar
        dest=/var/lib/ceph/
        copy=yes
        mode=0600
        creates=etc/ceph/ceph.conf
      when: migration_completed.stat.exists == False

    - name: Copy keys and configs
      shell: >
        cp etc/ceph/* /etc/ceph/
        chdir=/var/lib/ceph/
      when: migration_completed.stat.exists == False

    # NOTE (leseb): at this point the upstart and sysvinit checks are not necessary
    # so we directly call sysvinit
    - name: Start all the OSDs
      service: >
        name=ceph-osd-all
        state=started
        args=osd
      when: migration_completed.stat.exists == False

    # NOTE (leseb): this is tricky unless this is set into the ceph.conf
    # listened ports can be predicted, thus they will change after each restart
#    - name: Wait for the OSDs to be up again
#      local_action: >
#        wait_for
#        host={{ ansible_ssh_host | default(inventory_hostname) }}
#        port={{ item }}
#        timeout=30
#      with_items:
#        - "{{ osd_ports.stdout_lines }}"

    - name: Waiting for clean PGs...
      shell: >
        test "[""$(ceph -s -f json | python -c 'import sys, json; print(json.load(sys.stdin)["pgmap"]["num_pgs"])')""]" = "$(ceph -s -f json | python -c 'import sys, json; print([ i["count"] for i in json.load(sys.stdin)["pgmap"]["pgs_by_state"] if i["state_name"] == "active+clean"])')"
      register: result
      until: result.rc == 0
      retries: 10
      delay: 10
      delegate_to: "{{ item }}"
      with_items: "{{ groups.backup[0] }}"
      when: migration_completed.stat.exists == False

    - name: Done moving to the next OSD
      file: >
        path=/var/lib/ceph/migration_completed
        state=touch
        owner=root
        group=root
        mode=0600
      when: migration_completed.stat.exists == False

    - name: Unset the noout flag
      command: ceph osd unset noout
      delegate_to: "{{ item }}"
      with_items: "{{ groups[mon_group_name][0] }}"
      when: migration_completed.stat.exists == False

- hosts: rgws
  serial: 1
  sudo: True

  vars:
    backup_dir: /tmp/

  tasks:
    - name: Check if the node has be migrated already
      stat: >
        path=/var/lib/ceph/radosgw/migration_completed
      register: migration_completed
      failed_when: false

    - name: Check for failed run
      stat: >
        path=/var/lib/ceph/{{ ansible_facts['hostname'] }}.tar
      register: rgw_archive_leftover

    - fail: msg="Looks like an archive is already there, please remove it!"
      when: migration_completed.stat.exists == False and rgw_archive_leftover.stat.exists == True

    - name: Archive rados gateway configs
      shell: >
        tar -cpvzf - --one-file-system . /etc/ceph/* | cat > {{ ansible_facts['hostname'] }}.tar
        chdir=/var/lib/ceph/
        creates={{ ansible_facts['hostname'] }}.tar
      when: migration_completed.stat.exists == False

    - name: Create backup directory
      file: >
        path={{ backup_dir }}/rgws-backups
        state=directory
        owner=root
        group=root
        mode=0644
      delegate_to: "{{ item }}"
      with_items: "{{ groups.backup[0] }}"
      when: migration_completed.stat.exists == False

    - name: Scp RGWs dirs and configs
      fetch: >
        src=/var/lib/ceph/{{ ansible_facts['hostname'] }}.tar
        dest={{ backup_dir }}/rgws-backups/
        flat=yes
      when: migration_completed.stat.exists == False

    - name: Gracefully stop the rados gateway
      service: >
        name={{ item }}
        state=stopped
      with_items: radosgw
      when: migration_completed.stat.exists == False

    - name: Wait for radosgw to be down
      local_action:
        module: wait_for
        host: "{{ ansible_ssh_host | default(inventory_hostname) }}"
        path: /tmp/radosgw.sock
        state: absent
        timeout: 30
      when: migration_completed.stat.exists == False

    - name: Reboot the server
      command: reboot
      when: migration_completed.stat.exists == False

    - name: Wait for the server to come up
      local_action:
        module: wait_for
        port: 22
        delay: 10
        timeout: 3600
      when: migration_completed.stat.exists == False

    - name: Wait a bit to be sure that the server is ready for scp
      pause: seconds=20
      when: migration_completed.stat.exists == False

    # NOTE (leseb): 'creates' was added in Ansible 1.6
    - name: Copy and unarchive the OSD configs
      unarchive: >
        src={{ backup_dir }}/rgws-backups/{{ ansible_facts['hostname'] }}.tar
        dest=/var/lib/ceph/
        copy=yes
        mode=0600
        creates=etc/ceph/ceph.conf
      when: migration_completed.stat.exists == False

    - name: Copy keys and configs
      shell: >
        {{ item }}
        chdir=/var/lib/ceph/
      with_items: cp etc/ceph/* /etc/ceph/
      when: migration_completed.stat.exists == False

    - name: Start rados gateway
      service: >
        name={{ item }}
        state=started
      with_items: radosgw
      when: migration_completed.stat.exists == False

    - name: Wait for radosgw to be up again
      local_action:
        module: wait_for
        host: "{{ ansible_ssh_host | default(inventory_hostname) }}"
        path: /tmp/radosgw.sock
        state: present
        timeout: 30
      when: migration_completed.stat.exists == False

    - name: Done moving to the next rados gateway
      file: >
        path=/var/lib/ceph/radosgw/migration_completed
        state=touch
        owner=root
        group=root
        mode=0600
      when: migration_completed.stat.exists == False