fixed handling of new ceph_version

cluster-maintenance.yml

@@ -27,7 +27,7 @@
 
   - name: Wait for the server to come up
     local_action: >
-      wait_for host=<your_host
+      wait_for host=<your_host>
       port=22
       delay=10
       timeout=3600

group_vars/all.sample

@@ -184,11 +184,41 @@ dummy:
 #rbd_cache: "true"
 #rbd_cache_writethrough_until_flush: "true"
 #rbd_concurrent_management_ops: 20
+
 #rbd_client_directories: false # this will create rbd_client_log_path and rbd_client_admin_socket_path directories with proper permissions
-#rbd_client_directory_user: qemu
-#rbd_client_directory_group: libvirtd
-#rbd_client_log_file: /var/log/rbd-clients/qemu-guest-$pid.log # must be writable by QEMU and allowed by SELinux or AppArmor
+
+# Permissions for the rbd_client_log_path and
+# rbd_client_admin_socket_path. Depending on your use case for Ceph
+# you may want to change these values. The default, which is used if
+# any of the variables are unset or set to a false value (like `null`
+# or `false`) is to automatically determine what is appropriate for
+# the Ceph version with non-OpenStack workloads -- ceph:ceph and 0770
+# for infernalis releases, and root:root and 1777 for pre-infernalis
+# releases.
+#
+# For other use cases, including running Ceph with OpenStack, you'll
+# want to set these differently:
+#
+# For OpenStack on RHEL, you'll want:
+#   rbd_client_directory_owner: "qemu"
+#   rbd_client_directory_group: "libvirtd" (or "libvirt", depending on your version of libvirt)
+#   rbd_client_directory_mode: "0755"
+#
+# For OpenStack on Ubuntu or Debian, set:
+#    rbd_client_directory_owner: "libvirt-qemu"
+#    rbd_client_directory_group: "kvm"
+#    rbd_client_directory_mode: "0755"
+#
+# If you set rbd_client_directory_mode, you must use a string (e.g.,
+# 'rbd_client_directory_mode: "0755"', *not*
+# 'rbd_client_directory_mode: 0755', or Ansible will complain: mode
+# must be in octal or symbolic form
+#rbd_client_directory_owner: null
+#rbd_client_directory_group: null
+#rbd_client_directory_mode: null
+
 #rbd_client_log_path: /var/log/rbd-clients/
+#rbd_client_log_file: "{{ rbd_client_log_path }}/qemu-guest-$pid.log" # must be writable by QEMU and allowed by SELinux or AppArmor
 #rbd_client_admin_socket_path: /var/run/ceph/rbd-clients # must be writable by QEMU and allowed by SELinux or AppArmor
 #rbd_default_features: 3
 #rbd_default_map_options: rw

purge-cluster.yml

@@ -50,8 +50,8 @@
 # This can cause problem with qemu-kvm
     purge_all_packages: true
 
-# When set to true and raw _multi_journal is used then journal disk are also zapped
-    zap_journal_disks: true
+# When set to true and raw _multi_journal is used then block devices are also zapped
+    zap_block_devs: true
 
     ceph_packages:
       - ceph
@@ -96,6 +96,7 @@
     fail:
       msg: "OSD automatic discovery was detected, purge cluster does not support this scenario. If you want to purge the cluster, manually provide the list of devices in group_vars/osds using the devices variable."
     when:
+      osd_group_name in group_names and
       devices is not defined and
       osd_auto_discovery
 
@@ -108,7 +109,8 @@
     shell: "if [ -d /usr/lib/systemd ] ; then find /usr/lib/systemd/system -name 'ceph*' | wc -l ; else echo 0 ; fi"
     register: systemd_unit_files
 
-# Infernalis
+# after Hammer release
+
   - name: stop ceph.target with systemd
     service:
       name: ceph.target
@@ -166,34 +168,33 @@
       systemd_unit_files.stdout != "0" and
       rbdmirror_group_name in group_names
 
-# before infernalis
+# before infernalis release, using sysvinit scripts
+# we use this test so we do not have to know which RPM contains the boot script
+# or where it is placed.
+
   - name: stop ceph osds
-    command: service ceph stop osd
+    shell: "service ceph status osd ; if [ $? == 0 ] ; then service ceph stop osd ; else echo ; fi"
     when:
       ansible_os_family == 'RedHat' and
-      osd_group_name in group_names and
-      systemd_unit_files.stdout == "0"
+      osd_group_name in group_names
 
   - name: stop ceph mons
-    command: service ceph stop mon
+    shell: "service ceph status mon ; if [ $? == 0 ] ; then service ceph stop mon ; else echo ; fi"
     when:
       ansible_os_family == 'RedHat' and
-      mon_group_name in group_names and
-      systemd_unit_files.stdout == "0"
+      mon_group_name in group_names
 
   - name: stop ceph mdss
-    command: service ceph stop mds
+    shell: "service ceph status mds ; if [ $? == 0 ] ; then service ceph stop mds ; else echo ; fi"
     when:
       ansible_os_family == 'RedHat' and
-      mds_group_name in group_names and
-      systemd_unit_files.stdout == "0"
+      mds_group_name in group_names
 
   - name: stop ceph rgws
-    command: service ceph-radosgw stop
+    shell: "service ceph-radosgw status ; if [ $? == 0 ] ; then service ceph-radosgw stop ; else echo ; fi"
     when:
       ansible_os_family == 'RedHat' and
-      rgw_group_name in group_names and
-      systemd_unit_files.stdout == "0"
+      rgw_group_name in group_names
 
 # Ubuntu 14.04
   - name: stop ceph osds on ubuntu
@@ -240,6 +241,16 @@
     register: check_for_running_ceph
     failed_when: check_for_running_ceph.rc == 0
 
+  - name: see if ceph-disk-created data partitions are present
+    shell: "ls /dev/disk/by-partlabel | grep -q 'ceph\\\\x20data'"
+    failed_when: false
+    register: ceph_data_partlabels
+
+  - name: see if ceph-disk-created journal partitions are present
+    shell: "ls /dev/disk/by-partlabel | grep -q 'ceph\\\\x20journal'"
+    failed_when: false
+    register: ceph_journal_partlabels
+
   - name: get osd data mount points
     shell: "(grep /var/lib/ceph/osd /proc/mounts || echo -n) | awk '{ print $2 }'"
     register: mounted_osd
@@ -280,19 +291,29 @@
       osd_group_name in group_names and
       remove_osd_mountpoints.rc != 0
 
+  - name: see if ceph-disk is installed
+    shell: "which ceph-disk"
+    failed_when: false
+    register: ceph_disk_present
+
   - name: zap osd disks
     shell: ceph-disk zap "{{ item }}"
     with_items: devices
     when:
-     osd_group_name in group_names
+     osd_group_name in group_names and
+     ceph_disk_present.rc == 0 and
+     ceph_data_partlabels.rc == 0 and
+     zap_block_devs
 
   - name: zap journal devices
     shell: ceph-disk zap "{{ item }}"
     with_items: "{{ raw_journal_devices|default([])|unique }}"
     when:
       osd_group_name in group_names and
-      raw_multi_journal and
-      zap_journal_disks
+      ceph_disk_present.rc == 0 and
+      ceph_journal_partlabels.rc == 0 and
+      zap_block_devs and
+      raw_multi_journal
 
   - name: purge ceph packages with yum
     yum:

roles/ceph-common/defaults/main.yml

@@ -176,11 +176,41 @@ debug_mds_level: 20
 rbd_cache: "true"
 rbd_cache_writethrough_until_flush: "true"
 rbd_concurrent_management_ops: 20
+
 rbd_client_directories: false # this will create rbd_client_log_path and rbd_client_admin_socket_path directories with proper permissions
-rbd_client_directory_user: qemu
-rbd_client_directory_group: libvirtd
-rbd_client_log_file: /var/log/rbd-clients/qemu-guest-$pid.log # must be writable by QEMU and allowed by SELinux or AppArmor
+
+# Permissions for the rbd_client_log_path and
+# rbd_client_admin_socket_path. Depending on your use case for Ceph
+# you may want to change these values. The default, which is used if
+# any of the variables are unset or set to a false value (like `null`
+# or `false`) is to automatically determine what is appropriate for
+# the Ceph version with non-OpenStack workloads -- ceph:ceph and 0770
+# for infernalis releases, and root:root and 1777 for pre-infernalis
+# releases.
+#
+# For other use cases, including running Ceph with OpenStack, you'll
+# want to set these differently:
+#
+# For OpenStack on RHEL, you'll want:
+#   rbd_client_directory_owner: "qemu"
+#   rbd_client_directory_group: "libvirtd" (or "libvirt", depending on your version of libvirt)
+#   rbd_client_directory_mode: "0755"
+#
+# For OpenStack on Ubuntu or Debian, set:
+#    rbd_client_directory_owner: "libvirt-qemu"
+#    rbd_client_directory_group: "kvm"
+#    rbd_client_directory_mode: "0755"
+#
+# If you set rbd_client_directory_mode, you must use a string (e.g.,
+# 'rbd_client_directory_mode: "0755"', *not*
+# 'rbd_client_directory_mode: 0755', or Ansible will complain: mode
+# must be in octal or symbolic form
+rbd_client_directory_owner: null
+rbd_client_directory_group: null
+rbd_client_directory_mode: null
+
 rbd_client_log_path: /var/log/rbd-clients/
+rbd_client_log_file: "{{ rbd_client_log_path }}/qemu-guest-$pid.log" # must be writable by QEMU and allowed by SELinux or AppArmor
 rbd_client_admin_socket_path: /var/run/ceph/rbd-clients # must be writable by QEMU and allowed by SELinux or AppArmor
 rbd_default_features: 3
 rbd_default_map_options: rw

roles/ceph-common/tasks/installs/install_on_debian.yml

@@ -45,15 +45,3 @@
     default_release: "{{ ansible_distribution_release }}{{ '-backports' if ceph_origin == 'distro' and ceph_use_distro_backports else ''}}"
   when:
     mds_group_name in group_names
-
-- name: configure rbd clients directories
-  file:
-    path: "{{ item }}"
-    state: directory
-    owner: libvirt-qemu
-    group: kvm
-    mode: 0755
-  with_items:
-   - rbd_client_log_path
-   - rbd_client_admin_socket_path
-  when: rbd_client_directories

roles/ceph-common/tasks/installs/install_on_redhat.yml

@@ -139,15 +139,3 @@
   when:
     rgw_group_name in group_names and
     ansible_pkg_mgr == "dnf"
-
-- name: configure rbd clients directories
-  file:
-    path: "{{ item }}"
-    state: directory
-    owner: "{{ rbd_client_directory_user }}"
-    group: "{{ rbd_client_directory_group }}"
-    mode: 0755
-  with_items:
-   - rbd_client_log_path
-   - rbd_client_admin_socket_path
-  when: rbd_client_directories

roles/ceph-common/tasks/main.yml

@@ -115,16 +115,46 @@
   when: not is_ceph_infernalis
 
 - set_fact:
-    rbd_client_dir_owner: root
-    rbd_client_dir_group: root
-    rbd_client_dir_mode: "1777"
-  when: not is_ceph_infernalis
+    rbd_client_directory_owner: root
+  when:
+    not is_ceph_infernalis and
+    (rbd_client_directory_owner is not defined or
+     not rbd_client_directory_owner)
 
 - set_fact:
-    rbd_client_dir_owner: ceph
-    rbd_client_dir_group: ceph
-    rbd_client_dir_mode: "0770"
-  when: is_ceph_infernalis
+    rbd_client_directory_owner: ceph
+  when:
+    is_ceph_infernalis and
+    (rbd_client_directory_owner is not defined or
+     not rbd_client_directory_owner)
+
+- set_fact:
+    rbd_client_directory_group: root
+  when:
+    not is_ceph_infernalis and
+    (rbd_client_directory_group is not defined or
+     not rbd_client_directory_group)
+
+- set_fact:
+    rbd_client_directory_group: ceph
+  when:
+    is_ceph_infernalis and
+    (rbd_client_directory_group is not defined or
+     not rbd_client_directory_group)
+
+- set_fact:
+    rbd_client_directory_mode: "1777"
+  when:
+    not is_ceph_infernalis and
+    (rbd_client_directory_mode is not defined or
+     not rbd_client_directory_mode)
+
+- set_fact:
+    rbd_client_directory_mode: "0770"
+  when:
+    is_ceph_infernalis and
+    (rbd_client_directory_mode is not defined or
+     not rbd_client_directory_mode)
 
 - name: check for a ceph socket
   shell: "stat /var/run/ceph/*.asok > /dev/null 2>&1"
@@ -197,11 +227,15 @@
 
 - name: create rbd client directory
   file:
-    path: "{{ rbd_client_admin_socket_path }}"
+    path: "{{ item }}"
     state: directory
-    owner: "{{ rbd_client_dir_owner }}"
-    group: "{{ rbd_client_dir_group }}"
-    mode: "{{ rbd_client_dir_mode }}"
+    owner: "{{ rbd_client_directory_owner }}"
+    group: "{{ rbd_client_directory_group }}"
+    mode: "{{ rbd_client_directory_mode }}"
+  with_items:
+    - rbd_client_admin_socket_path
+    - rbd_client_log_path
+  when: rbd_client_directories
 
 - name: configure cluster name
   lineinfile:

roles/ceph-mds/tasks/docker/dirs_permissions.yml

@@ -7,7 +7,7 @@
 # NOTE (leseb): we can not use docker inspect with 'format filed' because of
 # https://github.com/ansible/ansible/issues/10156
 - name: inspect ceph version
-  shell: docker inspect docker.io/ceph/daemon | awk -F '=' '/CEPH_VERSION/ { gsub ("\",", "", $2); print $2 }' | uniq
+  shell: docker inspect docker.io/"{{ ceph_mds_docker_username }}"/"{{ ceph_mds_docker_imagename }}" | awk -F '=' '/CEPH_VERSION/ { gsub ("\",", "", $2); print $2 }' | uniq
   changed_when: false
   failed_when: false
   run_once: true

roles/ceph-mon/tasks/create_mds_filesystems.yml

@@ -9,9 +9,9 @@
     - cephfs_data
     - cephfs_metadata
   changed_when: false
-  when: not {{ ceph_version.stdout | version_compare('0.84', '<') }}
+  when: not {{ ceph_version | version_compare('0.80.0', '<') }}
 
 - name: create ceph filesystem
   command: ceph --cluster {{ cluster }} fs new {{ cephfs }} {{ cephfs_metadata }} {{ cephfs_data }}
   changed_when: false
-  when: not {{ ceph_version.stdout | version_compare('0.84', '<') }}
+  when: not {{ ceph_version | version_compare('0.80.0', '<') }}

roles/ceph-mon/tasks/docker/dirs_permissions.yml

@@ -7,7 +7,7 @@
 # NOTE (leseb): we can not use docker inspect with 'format filed' because of
 # https://github.com/ansible/ansible/issues/10156
 - name: inspect ceph version
-  shell: docker inspect docker.io/ceph/daemon | awk -F '=' '/CEPH_VERSION/ { gsub ("\",", "", $2); print $2 }' | uniq
+  shell: docker inspect docker.io/"{{ ceph_mon_docker_username }}"/"{{ ceph_mon_docker_imagename }}" | awk -F '=' '/CEPH_VERSION/ { gsub ("\",", "", $2); print $2 }' | uniq
   changed_when: false
   failed_when: false
   run_once: true

roles/ceph-mon/tasks/secure_cluster.yml

@@ -2,11 +2,11 @@
 - name: collect all the pools
   command: rados --cluster {{ cluster }} lspools
   register: ceph_pools
-  when: "{{ ceph_version.stdout | version_compare('0.94', '>=') }}"
+  when: "{{ ceph_version | version_compare('0.94.0', '>=') }}"
 
 - name: secure the cluster
   command: ceph --cluster {{ cluster }} osd pool set {{ item[0] }} {{ item[1] }} true
   with_nested:
     - ceph_pools.stdout_lines
     - secure_cluster_flags
-  when: "{{ ceph_version.stdout | version_compare('0.94', '>=') }}"
+  when: "{{ ceph_version | version_compare('0.94.0', '>=') }}"

roles/ceph-mon/tasks/start_monitor.yml

@@ -53,9 +53,3 @@
   changed_when: false
   failed_when:  false
   when: cephx
-
-- name: get ceph monitor version
-  shell: ceph --cluster {{ cluster }} daemon mon."{{ monitor_name }}" version | cut -d '"' -f 4 | cut -f 1,2 -d '.'
-  changed_when: false
-  failed_when: "'No such file or directory' in ceph_version.stderr"
-  register: ceph_version

roles/ceph-osd/tasks/activate_osds.yml

@@ -20,11 +20,19 @@
     - devices
   changed_when: false
   failed_when: false
+  register: activate_osd_disk
   when:
     not item.0.get("skipped") and
     item.0.get("rc", 0) != 0 and
     not osd_auto_discovery
 
+- name: fail if ceph-disk cannot create an OSD
+  fail:
+    msg: "ceph-disk failed to create an OSD"
+  when:
+    " 'ceph-disk: Error: ceph osd create failed' in item.get('stderr', '') "
+  with_items: "{{activate_osd_disk.results}}"
+
 # NOTE (leseb): this task is for partitions because we don't explicitly use a partition.
 - name: activate osd(s) when device is a partition
   command: "ceph-disk activate {{ item.1 }}"

roles/ceph-osd/tasks/docker/dirs_permissions.yml

@@ -7,7 +7,7 @@
 # NOTE (leseb): we can not use docker inspect with 'format filed' because of
 # https://github.com/ansible/ansible/issues/10156
 - name: inspect ceph version
-  shell: docker inspect docker.io/ceph/daemon | awk -F '=' '/CEPH_VERSION/ { gsub ("\",", "", $2); print $2 }' | uniq
+  shell: docker inspect docker.io/"{{ ceph_osd_docker_username }}"/"{{ ceph_osd_docker_imagename }}" | awk -F '=' '/CEPH_VERSION/ { gsub ("\",", "", $2); print $2 }' | uniq
   changed_when: false
   failed_when: false
   run_once: true

roles/ceph-restapi/tasks/docker/dirs_permissions.yml

@@ -1,6 +1,6 @@
 ---
 - name: inspect ceph version
-  shell: "docker inspect --format '{{ index (index .Config.Env) 3 }}' docker.io/{{ ceph_mon_docker_username }}/{{ ceph_mon_docker_imagename }} | cut -d '=' -f '2'"
+  shell: docker inspect docker.io/"{{ ceph_restapi_docker_username }}"/"{{ ceph_restapi_docker_imagename }}" | awk -F '=' '/CEPH_VERSION/ { gsub ("\",", "", $2); print $2 }' | uniq
   changed_when: false
   failed_when: false
   run_once: true

roles/ceph-rgw/tasks/docker/dirs_permissions.yml

@@ -7,7 +7,7 @@
 # NOTE (leseb): we can not use docker inspect with 'format filed' because of
 # https://github.com/ansible/ansible/issues/10156
 - name: inspect ceph version
-  shell: docker inspect docker.io/ceph/daemon | awk -F '=' '/CEPH_VERSION/ { gsub ("\",", "", $2); print $2 }' | uniq
+  shell: docker inspect docker.io/"{{ ceph_rgw_docker_username }}"/"{{ ceph_rgw_docker_imagename }}" | awk -F '=' '/CEPH_VERSION/ { gsub ("\",", "", $2); print $2 }' | uniq
   changed_when: false
   failed_when: false
   run_once: true

test.yml

@@ -8,6 +8,3 @@
     - ceph-mds
     - ceph-rgw
     - ceph-fetch-keys
-    - ceph-agent
-    - ceph-common-coreos
-    - ceph-rbd-mirror