Merge pull request #2100 from ceph/lvm-bluestore

ceph-volume lvm bluestore support
2017-10-27 17:36:16 +02:00 · 2017-10-27 17:36:16 +02:00 · faccd0acf0
parent 6ea92756c0 429622dd34
commit faccd0acf0
11 changed files with 315 additions and 33 deletions
--- a/docs/source/osds/scenarios.rst
+++ b/docs/source/osds/scenarios.rst
@ -170,14 +170,16 @@ is only available when the ceph release is Luminous or newer.
   The creation of the logical volumes is not supported by ``ceph-ansible``, ``ceph-volume``
   only creates OSDs from existing logical volumes.

-Use ``osd_scenario: lvm`` to enable this scenario. Currently we only support dedicated journals
-when using lvm, not collocated journals.

-To configure this scenario use the ``lvm_volumes`` config option. ``lvm_volumes``  is a list of dictionaries which can
-contain a ``data``, ``journal``, ``data_vg`` and ``journal_vg`` key. The ``data`` key represents the logical volume name that is to be used for your OSD
-data. The ``journal`` key represents the logical volume name, device or partition that will be used for your OSD journal. The ``data_vg``
-key represents the volume group name that your ``data`` logical volume resides on. This key is required for purging of OSDs created
-by this scenario. The ``journal_vg`` key is optional and should be the volume group name that your journal lv resides on, if applicable.
+``lvm_volumes`` is the config option that needs to be defined to configure the
+mappings for devices to be deployed. It is a list of dictionaries which expects
+a volume name and a volume group for logical volumes, but can also accept
+a device in the case of ``filestore`` for the ``journal``.
+
+The ``data`` key represents the logical volume name that is to be used for your
+OSD data.  The ``data_vg`` key represents the volume group name that your
+``data`` logical volume resides on. This key is required for purging of OSDs
+created by this scenario.

 .. note::
   Any logical volume or logical group used in ``lvm_volumes`` must be a name and not a path.
@ -185,8 +187,30 @@ by this scenario. The ``journal_vg`` key is optional and should be the volume gr
 .. note::
   You can not use the same journal for many OSDs.

+
+``filestore``
+^^^^^^^^^^^^^
+There is filestore support which can be enabled with::
+
+        osd_objectstore: filestore
+
+To configure this scenario use the ``lvm_volumes`` config option.
+``lvm_volumes``  is a list of dictionaries which expects a volume name and
+a volume group for logical volumes, but can also accept a device in the case of
+``filestore`` for the ``journal``.
+
+The following keys are accepted for a ``filestore`` deployment:
+
+* ``data``
+* ``data_vg``
+* ``journal``
+* ``journal_vg`` (not required if ``journal`` is a device and not a logical volume)
+
+The ``journal`` key represents the logical volume name, device or partition that will be used for your OSD journal.
+
 For example, a configuration to use the ``lvm`` osd scenario would look like::
-    
+
+    osd_objectstore: filestore
    osd_scenario: lvm
    lvm_volumes:
      - data: data-lv1
@ -199,3 +223,43 @@ For example, a configuration to use the ``lvm`` osd scenario would look like::
      - data: data-lv3
        journal: /dev/sdb1
        data_vg: vg2
+
+
+``bluestore``
+^^^^^^^^^^^^^
+This scenario allows a combination of devices to be used in an OSD.
+``bluestore`` can work just with a single "block" device (specified by the
+``data`` and ``data_vg``) or additionally with a ``block.wal`` and ``block.db``
+(interchangeably)
+
+The following keys are accepted for a ``bluestore`` deployment:
+
+* ``data`` (required)
+* ``data_vg`` (required)
+* ``db`` (optional for ``block.db``)
+* ``db_vg`` (optional for ``block.db``)
+* ``wal`` (optional for ``block.wal``)
+* ``wal_vg`` (optional for ``block.wal``)
+
+A ``bluestore`` lvm deployment, for all four different combinations supported
+could look like::
+
+    osd_objectstore: bluestore
+    osd_scenario: lvm
+    lvm_volumes:
+      - data: data-lv1
+        data_vg: vg1
+      - data: data-lv2
+        data_vg: vg1
+        wal: wal-lv1
+        wal_vg: vg2
+      - data: data-lv3
+        data_vg: vg2
+        db: db-lv1
+        db_vg: vg2
+      - data: data-lv4
+        data_vg: vg4
+        db: db-lv4
+        db_vg: vg4
+        wal: wal-lv4
+        wal_vg: vg4
--- a/group_vars/osds.yml.sample
+++ b/group_vars/osds.yml.sample
@ -197,14 +197,16 @@ dummy:
 #bluestore_wal_devices: "{{ dedicated_devices }}"

 # III. Use ceph-volume to create OSDs from logical volumes.
-# Use 'osd_scenario: lvm' to enable this scenario. Currently we only support dedicated journals
+# Use 'osd_scenario: lvm' to enable this scenario.
 # when using lvm, not collocated journals.
-# lvm_volumes is a list of dictionaries. Each dictionary must contain a data, journal and vg_name
-# key. Any logical volume or logical group used must be a name and not a path.
-# data must be a logical volume
-# journal can be either a lv, device or partition. You can not use the same journal for many data lvs.
+# lvm_volumes is a list of dictionaries.
+#
+# Filestore: Each dictionary must contain a data, journal and vg_name key. Any
+# logical volume or logical group used must be a name and not a path.  data
+# must be a logical volume. journal can be either a lv, device or partition.
+# You can not use the same journal for many data lvs.
 # data_vg must be the volume group name of the data lv
-# journal_vg is optional and must be the volume group name of the journal lv, if applicable
+# journal_vg is optional and must be the volume group name of the journal lv, if applicable.
 # For example:
 # lvm_volumes:
 #   - data: data-lv1
@ -217,6 +219,28 @@ dummy:
 #   - data: data-lv3
 #     journal: /dev/sdb1
 #     data_vg: vg2
+#
+# Bluestore: Each dictionary must contain at least data. When defining wal or
+# db, it must have both the lv name and vg group (db and wal are not required).
+# This allows for four combinations: just data, data and wal, data and wal and
+# db, data and db.
+# For example:
+# lvm_volumes:
+#   - data: data-lv1
+#     data_vg: vg1
+#     wal: wal-lv1
+#     wal_vg: vg1
+#   - data: data-lv2
+#     db: db-lv2
+#     db_vg: vg2
+#   - data: data-lv3
+#     wal: wal-lv1
+#     wal_vg: vg3
+#     db: db-lv3
+#     db_vg: vg3
+#   - data: data-lv4
+#     data_vg: vg4
+
 #lvm_volumes: []


--- a/roles/ceph-osd/defaults/main.yml
+++ b/roles/ceph-osd/defaults/main.yml
@ -189,14 +189,16 @@ dedicated_devices: []
 bluestore_wal_devices: "{{ dedicated_devices }}"

 # III. Use ceph-volume to create OSDs from logical volumes.
-# Use 'osd_scenario: lvm' to enable this scenario. Currently we only support dedicated journals
+# Use 'osd_scenario: lvm' to enable this scenario.
 # when using lvm, not collocated journals.
-# lvm_volumes is a list of dictionaries. Each dictionary must contain a data, journal and vg_name
-# key. Any logical volume or logical group used must be a name and not a path.
-# data must be a logical volume
-# journal can be either a lv, device or partition. You can not use the same journal for many data lvs.
+# lvm_volumes is a list of dictionaries.
+#
+# Filestore: Each dictionary must contain a data, journal and vg_name key. Any
+# logical volume or logical group used must be a name and not a path.  data
+# must be a logical volume. journal can be either a lv, device or partition.
+# You can not use the same journal for many data lvs.
 # data_vg must be the volume group name of the data lv
-# journal_vg is optional and must be the volume group name of the journal lv, if applicable
+# journal_vg is optional and must be the volume group name of the journal lv, if applicable.
 # For example:
 # lvm_volumes:
 #   - data: data-lv1
@ -209,6 +211,28 @@ bluestore_wal_devices: "{{ dedicated_devices }}"
 #   - data: data-lv3
 #     journal: /dev/sdb1
 #     data_vg: vg2
+#
+# Bluestore: Each dictionary must contain at least data. When defining wal or
+# db, it must have both the lv name and vg group (db and wal are not required).
+# This allows for four combinations: just data, data and wal, data and wal and
+# db, data and db.
+# For example:
+# lvm_volumes:
+#   - data: data-lv1
+#     data_vg: vg1
+#     wal: wal-lv1
+#     wal_vg: vg1
+#   - data: data-lv2
+#     db: db-lv2
+#     db_vg: vg2
+#   - data: data-lv3
+#     wal: wal-lv1
+#     wal_vg: vg3
+#     db: db-lv3
+#     db_vg: vg3
+#   - data: data-lv4
+#     data_vg: vg4
+
 lvm_volumes: []


--- a/roles/ceph-osd/tasks/check_mandatory_vars.yml
+++ b/roles/ceph-osd/tasks/check_mandatory_vars.yml
@ -56,16 +56,6 @@
    - osd_scenario == "lvm"
    - ceph_release_num[ceph_release] < ceph_release_num.luminous

- name: verify osd_objectstore is 'filestore' when using the lvm osd_scenario
-  fail:
-    msg: "the lvm osd_scenario currently only works for filestore, not bluestore"
-  when:
-    - osd_group_name is defined
-    - osd_group_name in group_names
-    - osd_scenario == "lvm"
-    - not osd_auto_discovery
-    - osd_objectstore != 'filestore'
-
 - name: verify lvm_volumes have been provided
  fail:
    msg: "please provide lvm_volumes to your osd scenario"
--- a/roles/ceph-osd/tasks/scenarios/lvm.yml
+++ b/roles/ceph-osd/tasks/scenarios/lvm.yml
@ -1,12 +1,80 @@
 ---
+
+- name: list all lvm osds
+  command: ceph-volume lvm list
+  register: ceph_volume_lvm_list
+  failed_when: False
+  changed_when: False
+  check_mode: no
+
 - name: use ceph-volume to create filestore osds with dedicated journals
-  command: "ceph-volume lvm create --filestore --data {{ item.data_vg }}/{{ item.data }} --journal {{ item.journal }}"
+  command: "ceph-volume --cluster {{ cluster }} lvm create --filestore --data {{ item.data_vg }}/{{ item.data }} --journal {{ item.journal }}"
+  environment:
+    CEPH_VOLUME_DEBUG: 1
  with_items: "{{ lvm_volumes }}"
  when:
    - item.journal_vg is not defined
+    - osd_objectstore == 'filestore'
+    - "'{{ item.data_vg }}/{{ item.data }}' not in ceph_volume_lvm_list.stdout"

 - name: use ceph-volume to create filestore osds with dedicated lv journals
-  command: "ceph-volume lvm create --filestore --data {{ item.data_vg }}/{{ item.data }} --journal {{item.journal_vg }}/{{ item.journal }}"
+  command: "ceph-volume --cluster {{ cluster }} lvm create --filestore --data {{ item.data_vg }}/{{ item.data }} --journal {{item.journal_vg }}/{{ item.journal }}"
+  environment:
+    CEPH_VOLUME_DEBUG: 1
  with_items: "{{ lvm_volumes }}"
  when:
    - item.journal_vg is defined
+    - osd_objectstore == 'filestore'
+    - "'{{ item.data_vg }}/{{ item.data }}' not in ceph_volume_lvm_list.stdout"
+
+- name: use ceph-volume to create bluestore osds with db and wal
+  command: "ceph-volume --cluster {{ cluster }} lvm create --bluestore --data {{ item.data_vg }}/{{ item.data }} --block.wal {{ item.wal_vg }}/{{ item.wal }} --block.db {{ item.db_vg }}/{{ item.db }}"
+  environment:
+    CEPH_VOLUME_DEBUG: 1
+  with_items: "{{ lvm_volumes }}"
+  when:
+    - osd_objectstore == 'bluestore'
+    - item.wal is defined
+    - item.wal_vg is defined
+    - item.db is defined
+    - item.db_vg is defined
+    - "'{{ item.data_vg }}/{{ item.data }}' not in ceph_volume_lvm_list.stdout"
+
+- name: use ceph-volume to create bluestore osds with db only
+  command: "ceph-volume --cluster {{ cluster }} lvm create --bluestore --data {{ item.data_vg }}/{{ item.data }} --block.db {{ item.db_vg }}/{{ item.db }}"
+  environment:
+    CEPH_VOLUME_DEBUG: 1
+  with_items: "{{ lvm_volumes }}"
+  when:
+    - osd_objectstore == 'bluestore'
+    - item.wal is not defined
+    - item.wal_vg is not defined
+    - item.db is defined
+    - item.db_vg is defined
+    - "'{{ item.data_vg }}/{{ item.data }}' not in ceph_volume_lvm_list.stdout"
+
+- name: use ceph-volume to create bluestore osds with wal only
+  command: "ceph-volume --cluster {{ cluster }} lvm create --bluestore --data {{ item.data_vg }}/{{ item.data }} --block.wal {{ item.wal_vg }}/{{ item.wal }}"
+  environment:
+    CEPH_VOLUME_DEBUG: 1
+  with_items: "{{ lvm_volumes }}"
+  when:
+    - osd_objectstore == 'bluestore'
+    - item.wal is defined
+    - item.wal_vg is defined
+    - item.db is not defined
+    - item.db_vg is not defined
+    - "'{{ item.data_vg }}/{{ item.data }}' not in ceph_volume_lvm_list.stdout"
+
+- name: use ceph-volume to create bluestore osds with just a data device
+  command: "ceph-volume --cluster {{ cluster }} lvm create --bluestore --data {{ item.data_vg }}/{{ item.data }}"
+  environment:
+    CEPH_VOLUME_DEBUG: 1
+  with_items: "{{ lvm_volumes }}"
+  when:
+    - osd_objectstore == 'bluestore'
+    - item.wal is not defined
+    - item.wal_vg is not defined
+    - item.db is not defined
+    - item.db_vg is not defined
+    - "'{{ item.data_vg }}/{{ item.data }}' not in ceph_volume_lvm_list.stdout"
--- a/tests/functional/centos/7/bs-lvm-osds/Vagrantfile
+++ b/tests/functional/centos/7/bs-lvm-osds/Vagrantfile
@ -0,0 +1 @@
+../../../../../Vagrantfile
--- a/tests/functional/centos/7/bs-lvm-osds/ceph-override.json
+++ b/tests/functional/centos/7/bs-lvm-osds/ceph-override.json
@ -0,0 +1 @@
+../cluster/ceph-override.json
--- a/tests/functional/centos/7/bs-lvm-osds/group_vars/all
+++ b/tests/functional/centos/7/bs-lvm-osds/group_vars/all
@ -0,0 +1,25 @@
+---
+
+ceph_origin: repository
+ceph_repository: community
+cluster: test
+public_network: "192.168.39.0/24"
+cluster_network: "192.168.40.0/24"
+monitor_interface: eth1
+osd_objectstore: "bluestore"
+osd_scenario: lvm
+copy_admin_key: true
+# test-volume is created by tests/functional/lvm_setup.yml from /dev/sdb
+lvm_volumes:
+  - data: data-lv1
+    data_vg: test_group
+  - data: data-lv2
+    data_vg: test_group
+    db: journal1
+    db_vg: journals
+os_tuning_params:
+  - { name: kernel.pid_max, value: 4194303 }
+  - { name: fs.file-max, value: 26234859 }
+ceph_conf_overrides:
+  global:
+    osd_pool_default_size: 1
--- a/tests/functional/centos/7/bs-lvm-osds/hosts
+++ b/tests/functional/centos/7/bs-lvm-osds/hosts
@ -0,0 +1,8 @@
+[mons]
+mon0
+
+[mgrs]
+mon0
+
+[osds]
+osd0
--- a/tests/functional/centos/7/bs-lvm-osds/vagrant_variables.yml
+++ b/tests/functional/centos/7/bs-lvm-osds/vagrant_variables.yml
@ -0,0 +1,74 @@
+---
+
+# DEPLOY CONTAINERIZED DAEMONS
+docker: false
+
+# DEFINE THE NUMBER OF VMS TO RUN
+mon_vms: 1
+osd_vms: 1
+mds_vms: 0
+rgw_vms: 0
+nfs_vms: 0
+rbd_mirror_vms: 0
+client_vms: 0
+iscsi_gw_vms: 0
+mgr_vms: 0
+
+# Deploy RESTAPI on each of the Monitors
+restapi: true
+
+# INSTALL SOURCE OF CEPH
+# valid values are 'stable' and 'dev'
+ceph_install_source: stable
+
+# SUBNETS TO USE FOR THE VMS
+public_subnet: 192.168.39
+cluster_subnet: 192.168.40
+
+# MEMORY
+# set 1024 for CentOS
+memory: 512
+
+# Ethernet interface name
+# use eth1 for libvirt and ubuntu precise, enp0s8 for CentOS and ubuntu xenial
+eth: 'eth1'
+
+# Disks
+# For libvirt use disks: "[ '/dev/vdb', '/dev/vdc' ]"
+# For CentOS7 use disks: "[ '/dev/sda', '/dev/sdb' ]"
+disks: "[ '/dev/sdb', '/dev/sdc' ]"
+
+# VAGRANT BOX
+# Ceph boxes are *strongly* suggested. They are under better control and will
+# not get updated frequently unless required for build systems. These are (for
+# now):
+#
+# * ceph/ubuntu-xenial
+#
+# Ubuntu: ceph/ubuntu-xenial bento/ubuntu-16.04 or ubuntu/trusty64 or ubuntu/wily64
+# CentOS: bento/centos-7.1 or puppetlabs/centos-7.0-64-puppet
+# libvirt CentOS: centos/7
+# parallels Ubuntu: parallels/ubuntu-14.04
+# Debian: deb/jessie-amd64 - be careful the storage controller is named 'SATA Controller'
+# For more boxes have a look at:
+#   - https://atlas.hashicorp.com/boxes/search?utf8=✓&sort=&provider=virtualbox&q=
+#   - https://download.gluster.org/pub/gluster/purpleidea/vagrant/
+vagrant_box: centos/7
+#ssh_private_key_path: "~/.ssh/id_rsa"
+# The sync directory changes based on vagrant box
+# Set to /home/vagrant/sync for Centos/7, /home/{ user }/vagrant for openstack and defaults to /vagrant
+#vagrant_sync_dir: /home/vagrant/sync
+vagrant_sync_dir: /vagrant
+# Disables synced folder creation. Not needed for testing, will skip mounting
+# the vagrant directory on the remote box regardless of the provider.
+vagrant_disable_synced_folder: true
+# VAGRANT URL
+# This is a URL to download an image from an alternate location.  vagrant_box
+# above should be set to the filename of the image.
+# Fedora virtualbox: https://download.fedoraproject.org/pub/fedora/linux/releases/22/Cloud/x86_64/Images/Fedora-Cloud-Base-Vagrant-22-20150521.x86_64.vagrant-virtualbox.box
+# Fedora libvirt: https://download.fedoraproject.org/pub/fedora/linux/releases/22/Cloud/x86_64/Images/Fedora-Cloud-Base-Vagrant-22-20150521.x86_64.vagrant-libvirt.box
+# vagrant_box_url: https://download.fedoraproject.org/pub/fedora/linux/releases/22/Cloud/x86_64/Images/Fedora-Cloud-Base-Vagrant-22-20150521.x86_64.vagrant-virtualbox.box
+
+os_tuning_params:
+  - { name: kernel.pid_max, value: 4194303 }
+  - { name: fs.file-max, value: 26234859 }
--- a/tox.ini
+++ b/tox.ini
@ -1,6 +1,6 @@
 [tox]
 envlist = {dev,jewel,luminous,rhcs}-{ansible2.2,ansible2.3}-{xenial_cluster,centos7_cluster,docker_cluster,update_cluster,cluster,update_docker_cluster,switch_to_containers,purge_filestore_osds_container,purge_filestore_osds_non_container,purge_cluster_non_container,purge_cluster_container}
-  {dev,luminous}-{ansible2.2,ansible2.3}-{filestore_osds_non_container,filestore_osds_container,bluestore_osds_container,bluestore_osds_non_container,lvm_osds,purge_lvm_osds,shrink_mon,shrink_osd,shrink_mon_container,shrink_osd_container,docker_cluster_collocation,purge_bluestore_osds_non_container,purge_bluestore_osds_container}
+  {dev,luminous}-{ansible2.2,ansible2.3}-{filestore_osds_non_container,filestore_osds_container,bluestore_osds_container,bluestore_osds_non_container,bluestore_lvm_osds,lvm_osds,purge_lvm_osds,shrink_mon,shrink_osd,shrink_mon_container,shrink_osd_container,docker_cluster_collocation,purge_bluestore_osds_non_container,purge_bluestore_osds_container}

 skipsdist = True

@ -151,6 +151,7 @@ setenv=
  luminous: UPDATE_CEPH_STABLE_RELEASE = luminous
  luminous: UPDATE_CEPH_DOCKER_IMAGE_TAG = tag-build-master-luminous-ubuntu-16.04
  lvm_osds: CEPH_STABLE_RELEASE = luminous
+  bluestore_lvm_osds: CEPH_STABLE_RELEASE = luminous
 deps=
  ansible2.2: ansible==2.2.3
  ansible2.3: ansible==2.3.1
@ -183,6 +184,7 @@ changedir=
  update_cluster: {toxinidir}/tests/functional/centos/7/cluster
  switch_to_containers: {toxinidir}/tests/functional/centos/7/cluster
  lvm_osds: {toxinidir}/tests/functional/centos/7/lvm-osds
+  bluestore_lvm_osds: {toxinidir}/tests/functional/centos/7/bs-lvm-osds
  purge_lvm_osds: {toxinidir}/tests/functional/centos/7/lvm-osds

 commands=
@ -193,6 +195,7 @@ commands=
  bash {toxinidir}/tests/scripts/generate_ssh_config.sh {changedir}

  lvm_osds: ansible-playbook -vv -i {changedir}/hosts {toxinidir}/tests/functional/lvm_setup.yml
+  bluestore_lvm_osds: ansible-playbook -vv -i {changedir}/hosts {toxinidir}/tests/functional/lvm_setup.yml
  purge_lvm_osds: ansible-playbook -vv -i {changedir}/hosts {toxinidir}/tests/functional/lvm_setup.yml

  rhcs: ansible-playbook -vv -i {changedir}/hosts {toxinidir}/tests/functional/rhcs_setup.yml --extra-vars "ceph_docker_registry={env:CEPH_DOCKER_REGISTRY:docker.io} repo_url={env:REPO_URL:} rhel7_repo_url={env:RHEL7_REPO_URL:}" --skip-tags "vagrant_setup"