Automate setting up a cache tier

Introduces a new flag to bootstrap cache devices

DocImpact
Partially-Implements: blueprint ceph-improvements

Change-Id: I09b5a0d5c61b3465237e5f01dc10120725561cd3

ansible/group_vars/all.yml

@@ -173,6 +173,12 @@ cinder_volume_driver: "{{ 'ceph' if enable_ceph | bool else 'lvm' }}"
 ###################
 # Ceph options
 ###################
+# Ceph can be setup with a caching to improve performance. To use the cache you
+# must provide seperate disks than those for the OSDs
+ceph_enable_cache: "no"
+# Valid options are [ forward, none, writeback ]
+ceph_cache_mode: "writeback"
+
 # A requirement for using the erasure-coded pools is you must setup a cache tier
 # Valid options are [ erasure, replicated ]
 ceph_pool_type: "replicated"
@@ -181,3 +187,7 @@ ceph_cinder_pool_name: "volumes"
 ceph_cinder_backup_pool_name: "backups"
 ceph_glance_pool_name: "images"
 ceph_nova_pool_name: "vms"
+
+ceph_erasure_profile: "k=4 m=2 ruleset-failure-domain=host"
+ceph_rule: "default host {{ 'indep' if ceph_pool_type == 'erasure' else 'firstn' }}"
+ceph_cache_rule: "cache host firstn"

ansible/roles/ceph/tasks/bootstrap_osds.yml

@@ -5,6 +5,12 @@
   register: osds_bootstrap
   when: inventory_hostname in groups['ceph-osd']
 
+- name: Looking up disks to bootstrap for Ceph cache
+  find_disks:
+      partition_name: 'KOLLA_CEPH_OSD_CACHE_BOOTSTRAP'
+  register: osds_cache_bootstrap
+  when: inventory_hostname in groups['ceph-osd']
+
 - name: Bootstrapping Ceph OSDs
   docker:
     tty: True
@@ -47,3 +53,47 @@
     state: absent
   with_indexed_items: osds_bootstrap['disks']|default([])
   when: inventory_hostname in groups['ceph-osd']
+
+- name: Bootstrapping Ceph Cache OSDs
+  docker:
+    tty: True
+    net: host
+    pull: "{{ docker_pull_policy }}"
+    restart_policy: "no"
+    state: reloaded
+    registry: "{{ docker_registry }}"
+    username: "{{ docker_registry_username }}"
+    password: "{{ docker_registry_password }}"
+    insecure_registry: "{{ docker_insecure_registry }}"
+    privileged: True
+    name: "bootstrap_osd_cache_{{ item.0 }}"
+    image: "{{ ceph_osd_image_full }}"
+    volumes:
+      - "{{ node_config_directory }}/ceph-osd/:{{ container_config_directory }}/:ro"
+      - "/dev/:/dev/"
+    env:
+      KOLLA_BOOTSTRAP:
+      KOLLA_CONFIG_STRATEGY: "{{ config_strategy }}"
+      CEPH_CACHE:
+      OSD_DEV: "{{ item.1.device }}"
+      OSD_INITIAL_WEIGHT: "{{ osd_initial_weight }}"
+  with_indexed_items: osds_cache_bootstrap['disks']|default([])
+  when: inventory_hostname in groups['ceph-osd']
+
+# https://github.com/ansible/ansible-modules-core/pull/1031
+- name: Waiting for bootstrap containers to exit
+  command: docker wait "bootstrap_osd_cache_{{ item.0 }}"
+  register: bootstrap_result
+  run_once: True
+  failed_when: bootstrap_result.stdout != "0"
+  with_indexed_items: osds_cache_bootstrap['disks']|default([])
+  when: inventory_hostname in groups['ceph-osd']
+
+- name: Cleaning up bootstrap containers
+  docker:
+    tty: True
+    name: "bootstrap_osd_cache_{{ item.0 }}"
+    image: "{{ ceph_osd_image_full }}"
+    state: absent
+  with_indexed_items: osds_cache_bootstrap['disks']|default([])
+  when: inventory_hostname in groups['ceph-osd']

ansible/roles/ceph_pools.yml

@@ -0,0 +1,80 @@
+---
+# TODO(SamYaple): Improve failed_when and changed_when tests if possible
+- name: Creating ceph erasure profile
+  command: docker exec ceph_mon ceph osd erasure-code-profile set erasure-profile {{ ceph_erasure_profile }}
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  failed_when: False
+  run_once: True
+  when: pool_type == "erasure"
+
+- name: Creating ceph ruleset
+  command: docker exec ceph_mon ceph osd crush rule create-erasure disks erasure-profile
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  failed_when: False
+  run_once: True
+  when: pool_type == "erasure"
+
+- name: Creating ceph ruleset
+  command: docker exec ceph_mon ceph osd crush rule create-simple disks {{ ceph_rule }}
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  failed_when: False
+  run_once: True
+  when: pool_type == "replicated"
+
+- name: Creating ceph pool
+  command: docker exec ceph_mon ceph osd pool create {{ pool_name }} 128 128 {{ pool_type }} {{ 'erasure-profile' if pool_type == 'erasure' else '' }} disks
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  failed_when: False
+  run_once: True
+
+- name: Creating ceph ruleset for cache
+  command: docker exec ceph_mon ceph osd crush rule create-simple cache {{ ceph_cache_rule }}
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  failed_when: False
+  run_once: True
+  when: "{{ ceph_enable_cache | bool }}"
+
+- name: Creating ceph pool for cache
+  command: docker exec ceph_mon ceph osd pool create {{ pool_name }}-cache 128 128 replicated cache
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  failed_when: False
+  run_once: True
+  when: "{{ ceph_enable_cache | bool }}"
+
+- name: Adding cache to pool
+  command: docker exec ceph_mon ceph osd tier add {{ pool_name }} {{ pool_name }}-cache
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  failed_when: False
+  run_once: True
+  when: "{{ ceph_enable_cache | bool }}"
+
+- name: Setting cache-mode
+  command: docker exec ceph_mon ceph osd tier cache-mode {{ pool_name }}-cache {{ cache_mode }}
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  failed_when: False
+  run_once: True
+  when: "{{ ceph_enable_cache | bool }}"
+
+- name: Setting cache overlay for pool
+  command: docker exec ceph_mon ceph osd tier set-overlay {{ pool_name }} {{ pool_name }}-cache
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  failed_when: False
+  run_once: True
+  when: "{{ ceph_enable_cache | bool }}"
+
+- name: Setting cache hit_set_type
+  command: docker exec ceph_mon ceph osd pool set {{ pool_name }}-cache hit_set_type bloom
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  failed_when: False
+  run_once: True
+  when: "{{ ceph_enable_cache | bool }}"

ansible/roles/cinder/defaults/main.yml

@@ -1,6 +1,24 @@
 ---
 project_name: "cinder"
 
+####################
+# Ceph
+####################
+ceph_cinder_pool_type: "{{ ceph_pool_type }}"
+ceph_cinder_cache_mode: "{{ ceph_cache_mode }}"
+ceph_cinder_backup_pool_type: "{{ ceph_pool_type }}"
+ceph_cinder_backup_cache_mode: "{{ ceph_cache_mode }}"
+
+# Due to Ansible issues on include, you cannot override these variables. Please
+# override the variables they refernce instead.
+cinder_pool_name: "{{ ceph_cinder_pool_name }}"
+cinder_pool_type: "{{ ceph_cinder_pool_type }}"
+cinder_cache_mode: "{{ ceph_cinder_cache_mode }}"
+cinder_backup_pool_name: "{{ ceph_cinder_backup_pool_name }}"
+cinder_backup_pool_type: "{{ ceph_cinder_backup_pool_type }}"
+cinder_backup_cache_mode: "{{ ceph_cinder_backup_cache_mode }}"
+
+
 ####################
 # Database
 ####################

ansible/roles/cinder/tasks/ceph.yml

@@ -17,21 +17,17 @@
     - "cinder-backup"
   when: inventory_hostname in groups['cinder-volume']
 
-# TODO(SamYaple): Improve failed_when and changed_when tests
-- name: Creating ceph pool for cinder
-  command: docker exec ceph_mon ceph osd pool create {{ ceph_cinder_pool_name }} 128 {{ ceph_pool_type }}
-  delegate_to: "{{ groups['ceph-mon'][0] }}"
-  changed_when: False
-  failed_when: False
-  run_once: True
+- include: ../../ceph_pools.yml
+  vars:
+    pool_name: "{{ cinder_pool_name }}"
+    pool_type: "{{ cinder_pool_type }}"
+    cache_mode: "{{ cinder_cache_mode }}"
 
-# TODO(SamYaple): Improve failed_when and changed_when tests
-- name: Creating ceph pool for cinder-backup
-  command: docker exec ceph_mon ceph osd pool create {{ ceph_cinder_backup_pool_name }} 128 {{ ceph_pool_type }}
-  delegate_to: "{{ groups['ceph-mon'][0] }}"
-  changed_when: False
-  failed_when: False
-  run_once: True
+- include: ../../ceph_pools.yml
+  vars:
+    pool_name: "{{ cinder_backup_pool_name }}"
+    pool_type: "{{ cinder_backup_pool_type }}"
+    cache_mode: "{{ cinder_backup_cache_mode }}"
 
 # TODO(SamYaple): Improve failed_when and changed_when tests
 - name: Pulling cephx keyring for cinder
@@ -40,6 +36,16 @@
   delegate_to: "{{ groups['ceph-mon'][0] }}"
   changed_when: False
   run_once: True
+  when: not ceph_enable_cache | bool
+
+# TODO(SamYaple): Improve failed_when and changed_when tests
+- name: Pulling cephx keyring for cinder
+  command: docker exec ceph_mon ceph auth get-or-create client.cinder mon 'allow r' osd 'allow class-read object_prefix rbd_children, allow rwx pool={{ ceph_cinder_pool_name }}, allow rwx pool={{ ceph_cinder_pool_name }}-cache, allow rwx pool={{ ceph_nova_pool_name }}, allow rwx pool={{ ceph_nova_pool_name }}-cache, allow rx pool={{ ceph_glance_pool_name }}, allow rx pool={{ ceph_glance_pool_name }}-cache'
+  register: cephx_key_cinder
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  run_once: True
+  when: ceph_enable_cache | bool
 
 # TODO(SamYaple): Improve failed_when and changed_when tests
 - name: Pulling cephx keyring for cinder-backup
@@ -48,6 +54,16 @@
   delegate_to: "{{ groups['ceph-mon'][0] }}"
   changed_when: False
   run_once: True
+  when: not ceph_enable_cache | bool
+
+# TODO(SamYaple): Improve failed_when and changed_when tests
+- name: Pulling cephx keyring for cinder-backup
+  command: docker exec ceph_mon ceph auth get-or-create client.cinder-backup mon 'allow r' osd 'allow class-read object_prefix rbd_children, allow rwx pool={{ ceph_cinder_backup_pool_name }}, allow rwx pool={{ ceph_cinder_backup_pool_name }}-cache'
+  register: cephx_key_cinder_backup
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  run_once: True
+  when: ceph_enable_cache | bool
 
 - name: Pushing cephx keyring
   copy:

ansible/roles/glance/defaults/main.yml

@@ -1,6 +1,19 @@
 ---
 project_name: "glance"
 
+####################
+# Ceph
+####################
+ceph_glance_pool_type: "{{ ceph_pool_type }}"
+ceph_glance_cache_mode: "{{ ceph_cache_mode }}"
+
+# Due to Ansible issues on include, you cannot override these variables. Please
+# override the variables they refernce instead.
+glance_pool_name: "{{ ceph_glance_pool_name }}"
+glance_pool_type: "{{ ceph_glance_pool_type }}"
+glance_cache_mode: "{{ ceph_glance_cache_mode }}"
+
+
 ####################
 # Database
 ####################

ansible/roles/glance/tasks/ceph.yml

@@ -11,13 +11,11 @@
     dest: "{{ node_config_directory }}/glance-api/ceph.conf"
   when: inventory_hostname in groups['glance-api']
 
-# TODO(SamYaple): Improve failed_when and changed_when tests
-- name: Creating ceph pool for glance
-  command: docker exec ceph_mon ceph osd pool create {{ ceph_glance_pool_name }} 128 {{ ceph_pool_type }}
-  delegate_to: "{{ groups['ceph-mon'][0] }}"
-  changed_when: False
-  failed_when: False
-  run_once: True
+- include: ../../ceph_pools.yml
+  vars:
+    pool_name: "{{ glance_pool_name }}"
+    pool_type: "{{ glance_pool_type }}"
+    cache_mode: "{{ glance_cache_mode }}"
 
 # TODO(SamYaple): Improve failed_when and changed_when tests
 - name: Pulling cephx keyring
@@ -26,6 +24,16 @@
   delegate_to: "{{ groups['ceph-mon'][0] }}"
   changed_when: False
   run_once: True
+  when: not ceph_enable_cache | bool
+
+# TODO(SamYaple): Improve failed_when and changed_when tests
+- name: Pulling cephx keyring
+  command: docker exec ceph_mon ceph auth get-or-create client.glance mon 'allow r' osd 'allow class-read object_prefix rbd_children, allow rwx pool={{ ceph_glance_pool_name }}, allow rwx pool={{ ceph_glance_pool_name }}-cache'
+  register: cephx_key
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  run_once: True
+  when: ceph_enable_cache | bool
 
 - name: Pushing cephx keyring
   copy:

ansible/roles/nova/defaults/main.yml

@@ -1,6 +1,19 @@
 ---
 project_name: "nova"
 
+####################
+# Ceph
+####################
+ceph_nova_pool_type: "{{ ceph_pool_type }}"
+ceph_nova_cache_mode: "{{ ceph_cache_mode }}"
+
+# Due to Ansible issues on include, you cannot override these variables. Please
+# override the variables they refernce instead.
+nova_pool_name: "{{ ceph_nova_pool_name }}"
+nova_pool_type: "{{ ceph_nova_pool_type }}"
+nova_cache_mode: "{{ ceph_nova_cache_mode }}"
+
+
 ####################
 # Database
 ####################

ansible/roles/nova/tasks/ceph.yml

@@ -17,21 +17,29 @@
     - "nova-libvirt"
   when: inventory_hostname in groups['compute']
 
-# TODO(SamYaple): Improve failed_when and changed_when tests
-- name: Creating ceph pool for vms
-  command: docker exec ceph_mon ceph osd pool create {{ ceph_nova_pool_name }} 128 {{ ceph_pool_type }}
-  delegate_to: "{{ groups['ceph-mon'][0] }}"
-  changed_when: False
-  failed_when: False
-  run_once: True
+- include: ../../ceph_pools.yml
+  vars:
+    pool_name: "{{ nova_pool_name }}"
+    pool_type: "{{ nova_pool_type }}"
+    cache_mode: "{{ nova_cache_mode }}"
 
 # TODO(SamYaple): Improve failed_when and changed_when tests
 - name: Pulling cephx keyring for nova
-  command: docker exec ceph_mon ceph auth get-or-create client.nova mon 'allow r' osd 'allow class-read object_prefix rbd_children, allow rwx pool={{ ceph_volumes_pool_name }}, allow rwx pool={{ ceph_nova_pool_name }}, allow rx pool={{ ceph_glance_pool_name }}'
+  command: docker exec ceph_mon ceph auth get-or-create client.nova mon 'allow r' osd 'allow class-read object_prefix rbd_children, allow rwx pool={{ ceph_cinder_pool_name }}, allow rwx pool={{ ceph_nova_pool_name }}, allow rx pool={{ ceph_glance_pool_name }}'
   register: cephx_key
   delegate_to: "{{ groups['ceph-mon'][0] }}"
   changed_when: False
   run_once: True
+  when: not ceph_enable_cache | bool
+
+# TODO(SamYaple): Improve failed_when and changed_when tests
+- name: Pulling cephx keyring for nova
+  command: docker exec ceph_mon ceph auth get-or-create client.nova mon 'allow r' osd 'allow class-read object_prefix rbd_children, allow rwx pool={{ ceph_cinder_pool_name }}, allow rwx pool={{ ceph_cinder_pool_name }}-cache, allow rwx pool={{ ceph_nova_pool_name }}, allow rwx pool={{ ceph_nova_pool_name }}-cache, allow rx pool={{ ceph_glance_pool_name }}, allow rx pool={{ ceph_glance_pool_name }}-cache'
+  register: cephx_key
+  delegate_to: "{{ groups['ceph-mon'][0] }}"
+  changed_when: False
+  run_once: True
+  when: ceph_enable_cache | bool
 
 # TODO(SamYaple): Improve failed_when and changed_when tests
 - name: Pulling cephx keyring for libvirt

docker/ceph/ceph-osd/extend_start.sh

@@ -26,13 +26,17 @@ if [[ "${!KOLLA_BOOTSTRAP[@]}" ]]; then
     ceph auth add "osd.${OSD_ID}" osd 'allow *' mon 'allow profile osd' -i "${OSD_DIR}/keyring"
     umount "${OSD_PARTITION}"
 
+    if [[ "${!CEPH_CACHE[@]}" ]]; then
+        CEPH_ROOT_NAME=cache
+    fi
+
     # These commands only need to be run once per host but are safe to run
     # repeatedly. This can be improved later or if any problems arise.
-    ceph osd crush add-bucket "$(hostname)" host
-    ceph osd crush move "$(hostname)" root=default
+    ceph osd crush add-bucket "$(hostname)${CEPH_ROOT_NAME:+-${CEPH_ROOT_NAME}}" host
+    ceph osd crush move "$(hostname)${CEPH_ROOT_NAME:+-${CEPH_ROOT_NAME}}" root=${CEPH_ROOT_NAME:-default}
 
     # Adding osd to crush map
-    ceph osd crush add "${OSD_ID}" "${OSD_INITIAL_WEIGHT}" host="$(hostname)"
+    ceph osd crush add "${OSD_ID}" "${OSD_INITIAL_WEIGHT}" host="$(hostname)${CEPH_ROOT_NAME:+-${CEPH_ROOT_NAME}}"
     exit 0
 fi
 

etc/kolla/globals.yml

@@ -89,6 +89,12 @@ neutron_external_interface: "eth1"
 ###################
 # Ceph options
 ###################
+# Ceph can be setup with a caching to improve performance. To use the cache you
+# must provide seperate disks than those for the OSDs
+# ceph_use_cache: "no"
+# Valid options are [ forward, none, writeback ]
+# ceph_cache_mode: writeback
+
 # A requirement for using the erasure-coded pools is you must setup a cache tier
 # Valid options are [ erasure, replicated ]
 # ceph_pool_type: "replicated"