shrink mon and osd

Rework shrinking a monitor and an OSD playbook. Also adding test
scenario.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1366807
Signed-off-by: Sébastien Han <seb@redhat.com>

Vagrantfile

@@ -187,7 +187,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
 
   (0..MGRS - 1).each do |i|
     config.vm.define "#{LABEL_PREFIX}mgr#{i}" do |mgr|
-      mgr.vm.hostname = "#{LABEL_PREFIX}ceph-mgr#{i}"
+      mgr.vm.hostname = "#{LABEL_PREFIX}mgr#{i}"
       if ASSIGN_STATIC_IP
         mgr.vm.network :private_network,
           ip: "#{PUBLIC_SUBNET}.3#{i}"
@@ -223,7 +223,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
   (0..CLIENTS - 1).each do |i|
     config.vm.define "#{LABEL_PREFIX}client#{i}" do |client|
       client.vm.box = CLIENT_BOX
-      client.vm.hostname = "#{LABEL_PREFIX}ceph-client#{i}"
+      client.vm.hostname = "#{LABEL_PREFIX}client#{i}"
       if ASSIGN_STATIC_IP
         client.vm.network :private_network,
           ip: "#{PUBLIC_SUBNET}.4#{i}"
@@ -258,7 +258,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
 
   (0..NRGWS - 1).each do |i|
     config.vm.define "#{LABEL_PREFIX}rgw#{i}" do |rgw|
-      rgw.vm.hostname = "#{LABEL_PREFIX}ceph-rgw#{i}"
+      rgw.vm.hostname = "#{LABEL_PREFIX}rgw#{i}"
       if ASSIGN_STATIC_IP
         rgw.vm.network :private_network,
           ip: "#{PUBLIC_SUBNET}.5#{i}"
@@ -294,7 +294,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
 
   (0..NNFSS - 1).each do |i|
     config.vm.define "nfs#{i}" do |nfs|
-      nfs.vm.hostname = "ceph-nfs#{i}"
+      nfs.vm.hostname = "#{LABEL_PREFIX}ceph-nfs#{i}"
       if ASSIGN_STATIC_IP
         nfs.vm.network :private_network,
           ip: "#{PUBLIC_SUBNET}.6#{i}"
@@ -330,7 +330,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
 
   (0..NMDSS - 1).each do |i|
     config.vm.define "#{LABEL_PREFIX}mds#{i}" do |mds|
-      mds.vm.hostname = "#{LABEL_PREFIX}ceph-mds#{i}"
+      mds.vm.hostname = "#{LABEL_PREFIX}mds#{i}"
       if ASSIGN_STATIC_IP
         mds.vm.network :private_network,
           ip: "#{PUBLIC_SUBNET}.7#{i}"
@@ -364,7 +364,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
 
   (0..NRBD_MIRRORS - 1).each do |i|
     config.vm.define "#{LABEL_PREFIX}rbd_mirror#{i}" do |rbd_mirror|
-      rbd_mirror.vm.hostname = "#{LABEL_PREFIX}ceph-rbd-mirror#{i}"
+      rbd_mirror.vm.hostname = "#{LABEL_PREFIX}rbd-mirror#{i}"
       if ASSIGN_STATIC_IP
         rbd_mirror.vm.network :private_network,
           ip: "#{PUBLIC_SUBNET}.8#{i}"
@@ -398,7 +398,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
 
   (0..NISCSI_GWS - 1).each do |i|
     config.vm.define "#{LABEL_PREFIX}iscsi_gw#{i}" do |iscsi_gw|
-      iscsi_gw.vm.hostname = "#{LABEL_PREFIX}ceph-iscsi-gw#{i}"
+      iscsi_gw.vm.hostname = "#{LABEL_PREFIX}iscsi-gw#{i}"
       if ASSIGN_STATIC_IP
         iscsi_gw.vm.network :private_network,
           ip: "#{PUBLIC_SUBNET}.9#{i}"
@@ -432,7 +432,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
 
   (0..NMONS - 1).each do |i|
     config.vm.define "#{LABEL_PREFIX}mon#{i}" do |mon|
-      mon.vm.hostname = "#{LABEL_PREFIX}ceph-mon#{i}"
+      mon.vm.hostname = "#{LABEL_PREFIX}mon#{i}"
       if ASSIGN_STATIC_IP
         mon.vm.network :private_network,
           ip: "#{PUBLIC_SUBNET}.1#{i}"
@@ -467,7 +467,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
 
   (0..NOSDS - 1).each do |i|
     config.vm.define "#{LABEL_PREFIX}osd#{i}" do |osd|
-      osd.vm.hostname = "#{LABEL_PREFIX}ceph-osd#{i}"
+      osd.vm.hostname = "#{LABEL_PREFIX}osd#{i}"
       if ASSIGN_STATIC_IP
         osd.vm.network :private_network,
           ip: "#{PUBLIC_SUBNET}.10#{i}"

infrastructure-playbooks/shrink-mon.yml

@@ -1,9 +1,9 @@
 ---
 # This playbook shrinks the Ceph monitors from your cluster
-# It can remove any number of monitor(s) from the cluster and ALL THEIR DATA
+# It can remove a Ceph of monitor from the cluster and ALL ITS DATA
 #
 # Use it like this:
-# ansible-playbook shrink-mon.yml -e mon_host=ceph-mon01,ceph-mon02
+# ansible-playbook shrink-mon.yml -e mon_to_kill=ceph-mon01
 #     Prompts for confirmation to shrink, defaults to no and
 #     doesn't shrink the cluster. yes shrinks the cluster.
 #
@@ -12,12 +12,21 @@
 #     automation scripts to avoid interactive prompt.
 
 
-- name: confirm whether user really meant to remove monitor(s) from the ceph cluster
+- name: gather facts and check the init system
+
+  hosts:
+    - "{{ mon_group_name|default('mons') }}"
+
+  become: true
+
+  tasks:
+    - debug: msg="gather facts on all Ceph hosts for following reference"
+
+- name: confirm whether user really meant to remove monitor from the ceph cluster
 
   hosts:
     - localhost
 
-  gather_facts: false
   become: true
 
   vars_prompt:
@@ -26,123 +35,110 @@
       default: 'no'
       private: no
 
-  tasks:
+  vars:
-  - include_vars: roles/ceph-defaults/defaults/main.yml
+    mon_group_name: mons
-  - include_vars: group_vars/all.yml
 
+  pre_tasks:
     - name: exit playbook, if only one monitor is present in cluster
       fail:
         msg: "You are about to shrink the only monitor present in the cluster.
               If you really want to do that, please use the purge-cluster playbook."
-    when: "{{ groups[mon_group_name] | length | int == 1 }}"
+      when:
+        - groups[mon_group_name] | length | int == 1
+
+    - name: exit playbook, if no monitor was given
+      fail:
+        msg: "mon_to_kill must be declared
+          Exiting shrink-cluster playbook, no monitor was removed.
+           On the command line when invoking the playbook, you can use
+           -e mon_to_kill=ceph-mon01 argument. You can only remove a single monitor each time the playbook runs."
+      when:
+        - mon_to_kill is not defined
+
+    - name: exit playbook, if the monitor is not part of the inventory
+      fail:
+        msg: "It seems that the host given is not part of your inventory, please make sure it is."
+      when:
+        - mon_to_kill not in groups[mon_group_name]
 
     - name: exit playbook, if user did not mean to shrink cluster
       fail:
-      msg: "Exiting shrink-mon playbook, no monitor(s) was/were removed.
+        msg: "Exiting shrink-mon playbook, no monitor was removed.
            To shrink the cluster, either say 'yes' on the prompt or
            or use `-e ireallymeanit=yes` on the command line when
            invoking the playbook"
-    when: ireallymeanit != 'yes'
-
-  - name: exit playbook, if no monitor(s) was/were given
-    fail:
-      msg: "mon_host must be declared
-        Exiting shrink-cluster playbook, no monitor(s) was/were removed.
-         On the command line when invoking the playbook, you can use
-         -e mon_host=ceph-mon01,ceph-mon02 argument."
-    when: mon_host is not defined
-
-  - name: test if ceph command exist
-    command: command -v ceph
-    changed_when: false
-    failed_when: false
-    register: ceph_command
-
-  - name: exit playbook, if ceph command does not exist
-    debug:
-      msg: "The ceph command is not available, please install it :("
-    run_once: true
       when:
-      - ceph_command.rc != 0
+        - ireallymeanit != 'yes'
 
-  - name: exit playbook, if cluster files do not exist
+  roles:
-    stat:
+    - ceph-defaults
-      path: "{{ item }}"
-    register: ceph_conf_key
-    with_items:
-      - /etc/ceph/{{ cluster }}.conf
-      - /etc/ceph/{{ cluster }}.client.admin.keyring
-    failed_when: false
 
-  - fail:
+  post_tasks:
-      msg: "Ceph's configuration file is not present in /etc/ceph"
+    - name: pick a monitor different than the one we want to remove
-    with_items: "{{ceph_conf_key.results}}"
+      set_fact: mon_host={{ item }}
+      with_items: "{{ groups[mon_group_name] }}"
       when:
-      -  item.stat.exists == false
+        - item != mon_to_kill
 
     - name: exit playbook, if can not connect to the cluster
       command: timeout 5 ceph --cluster {{ cluster }} health
       register: ceph_health
       until: ceph_health.stdout.find("HEALTH") > -1
+      delegate_to: "{{ mon_host }}"
       retries: 5
       delay: 2
 
-  - name: verify given monitors are reachable
+    - set_fact:
-    command: ping -c 1 {{ item }}
+        mon_to_kill_hostname: "{{ hostvars[mon_to_kill]['ansible_hostname'] }}"
-    with_items: "{{mon_host.split(',')}}"
-    register: mon_reachable
-    failed_when: false
 
-  - fail:
+    - name: stop monitor service(s)
-      msg: "One or more monitors are not reachable, please check your /etc/hosts or your DNS"
-    with_items: "{{mon_reachable.results}}"
-    when:
-      -  item.rc != 0
-
-  - name: stop monitor service
       service:
-      name: ceph-mon@{{ item }}
+        name: ceph-mon@{{ mon_to_kill_hostname }}
         state: stopped
         enabled: no
-    with_items: "{{mon_host.split(',')}}"
+      delegate_to: "{{ mon_to_kill }}"
-    delegate_to: "{{item}}"
       failed_when: false
 
     - name: purge monitor store
       file:
-      path: /var/lib/ceph/mon/{{ cluster }}-{{ item }}
+        path: /var/lib/ceph/mon/{{ cluster }}-{{ mon_to_kill_hostname }}
         state: absent
-    with_items: "{{mon_host.split(',')}}"
+      delegate_to: "{{ mon_to_kill }}"
-    delegate_to: "{{item}}"
 
     - name: remove monitor from the quorum
-    command: ceph --cluster {{ cluster }} mon remove {{ item }}
+      command: ceph --cluster {{ cluster }} mon remove {{ mon_to_kill_hostname }}
       failed_when: false
-    with_items: "{{mon_host.split(',')}}"
+      delegate_to: "{{ mon_host }}"
 
     # NOTE (leseb): sorry for the 'sleep' command
     # but it will take a couple of seconds for other monitors
     # to notice that one member has left.
     # 'sleep 5' is not that bad and should be sufficient
     - name: verify the monitor is out of the cluster
-    shell: "sleep 5 && ceph --cluster {{ cluster }} -s | grep monmap | sed 's/.*quorum//' | egrep -sq {{ item }}"
+      shell: |
-    with_items: "{{mon_host.split(',')}}"
+        ceph --cluster {{ cluster }} -s -f json | python -c 'import sys, json; print(json.load(sys.stdin)["quorum_names"])'
-    failed_when: false
+      delegate_to: "{{ mon_host }}"
-    register: ceph_health_mon
+      register: result
 
     - name: please remove the monitor from your ceph configuration file
       debug:
-        msg: "The monitor(s) has/have been successfully removed from the cluster.
+          msg: "The monitor has been successfully removed from the cluster.
-        Please remove the monitor(s) entry(ies) from the rest of your ceph configuration files, cluster wide."
+          Please remove the monitor entry from the rest of your ceph configuration files, cluster wide."
       run_once: true
-    with_items: "{{ceph_health_mon.results}}"
       when:
-      - item.rc != 0
+        - mon_to_kill_hostname not in result.stdout
 
-  - name: please remove the monitor from your ceph configuration file
+    - name: fail if monitor is still part of the cluster
       fail:
-        msg: "Monitor(s) appear(s) to still be part of the cluster, please check what happened."
+          msg: "Monitor appears to still be part of the cluster, please check what happened."
       run_once: true
-    with_items: "{{ceph_health_mon.results}}"
       when:
-      - item.rc == 0
+        - mon_to_kill_hostname in result.stdout
+
+    - name: show ceph health
+      command: ceph --cluster {{ cluster }} -s
+      delegate_to: "{{ mon_host }}"
+
+    - name: show ceph mon status
+      command: ceph --cluster {{ cluster }} mon stat
+      delegate_to: "{{ mon_host }}"

infrastructure-playbooks/shrink-osd.yml

@@ -3,7 +3,7 @@
 # It can remove any number of OSD(s) from the cluster and ALL THEIR DATA
 #
 # Use it like this:
-# ansible-playbook shrink-osd.yml -e osd_id=0,2,6
+# ansible-playbook shrink-osd.yml -e osd_to_kill=0,2,6
 #     Prompts for confirmation to shrink, defaults to no and
 #     doesn't shrink the cluster. yes shrinks the cluster.
 #
@@ -11,13 +11,21 @@
 #     Overrides the prompt using -e option. Can be used in
 #     automation scripts to avoid interactive prompt.
 
+- name: gather facts and check the init system
+
+  hosts:
+    - "{{ mon_group_name|default('mons') }}"
+    - "{{ osd_group_name|default('osds') }}"
+
+  become: True
+  tasks:
+    - debug: msg="gather facts on all Ceph hosts for following reference"
 
 - name: confirm whether user really meant to remove osd(s) from the cluster
 
   hosts:
     - localhost
 
-  gather_facts: false
   become: true
 
   vars_prompt:
@@ -26,10 +34,11 @@
       default: 'no'
       private: no
 
-  tasks:
+  vars:
-  - include_vars: roles/ceph-defaults/defaults/main.yml
+    mon_group_name: mons
-  - include_vars: group_vars/all.yml
+    osd_group_name: osds
 
+  pre_tasks:
     - name: exit playbook, if user did not mean to shrink cluster
       fail:
         msg: "Exiting shrink-osd playbook, no osd(s) was/were removed..
@@ -40,73 +49,40 @@
 
     - name: exit playbook, if no osd(s) was/were given
       fail:
-      msg: "osd_ids must be declared
+        msg: "osd_to_kill must be declared
           Exiting shrink-osd playbook, no OSD()s was/were removed.
            On the command line when invoking the playbook, you can use
-         -e osd_ids=0,1,2,3 argument."
+           -e osd_to_kill=0,1,2,3 argument."
-    when: osd_ids is not defined
+      when: osd_to_kill is not defined
 
-  - name: test if ceph command exist
+  roles:
-    command: command -v ceph
+    - ceph-defaults
-    changed_when: false
-    failed_when: false
-    register: ceph_command
 
-  - name: exit playbook, if ceph command does not exist
+  post_tasks:
-    debug:
-      msg: "The ceph command is not available, please install it :("
-    run_once: true
-    when:
-      - ceph_command.rc != 0
-
-  - name: exit playbook, if cluster files do not exist
-    stat:
-      path: "{{ item }}"
-    register: ceph_conf_key
-    with_items:
-      - /etc/ceph/{{ cluster }}.conf
-      - /etc/ceph/{{ cluster }}.client.admin.keyring
-    failed_when: false
-
-  - fail:
-      msg: "Ceph's configuration file is not present in /etc/ceph"
-    with_items: "{{ ceph_conf_key.results }}"
-    when:
-      -  item.stat.exists == false
 
     - name: exit playbook, if can not connect to the cluster
       command: timeout 5 ceph --cluster {{ cluster }} health
       register: ceph_health
       until: ceph_health.stdout.find("HEALTH") > -1
+      delegate_to: "{{ groups[mon_group_name][0] }}"
       retries: 5
       delay: 2
 
-# NOTE (leseb): just in case, the complex filters mechanism below does not work anymore.
+    - name: find the host(s) where the osd(s) is/are running on
-# This will be a quick and easy fix but will require using the shell module.
-#  - name: find the host where the osd(s) is/are running on
-#    shell: |
-#      ceph --cluster {{ cluster }} osd find {{ item }} | grep -Po '(?<="ip": ")[^:]*'
-#    with_items: "{{osd_ids.split(',')}}"
-#    register: osd_hosts
-#
-  - name: find the host where the osd(s) is/are running on
       command: ceph --cluster {{ cluster }} osd find {{ item }}
-    with_items: "{{ osd_ids.split(',') }}"
+      with_items: "{{ osd_to_kill.split(',') }}"
-    register: osd_hosts
+      delegate_to: "{{ groups[mon_group_name][0] }}"
+      register: find_osd_hosts
 
-  - set_fact: ip_item="{{(item.stdout | from_json).ip}}"
+    - set_fact:
-    with_items: "{{ osd_hosts.results }}"
+        osd_hosts: "{{ (item.stdout | from_json).crush_location.host }}"
-    register: ip_result
+      with_items: "{{ find_osd_hosts.results }}"
-
-  - set_fact: ips="{{ ip_result.results | map(attribute='ansible_facts.ip_item') | list }}"
-
-  - set_fact: real_ips="{{ ips | regex_replace(':[0-9][0-9][0-9][0-9]\/[0-9]+', '') }}"
 
     - name: check if ceph admin key exists on the osd nodes
       stat:
         path: "/etc/ceph/{{ cluster }}.client.admin.keyring"
       register: ceph_admin_key
-    with_items: "{{ real_ips }}"
+      with_items: "{{ osd_hosts }}"
       delegate_to: "{{ item }}"
       failed_when: false
 
@@ -121,14 +97,15 @@
       register: deactivate
       ignore_errors: yes
       with_together:
-      - "{{ osd_ids.split(',') }}"
+        - "{{ osd_to_kill.split(',') }}"
-      - "{{ real_ips }}"
+        - "{{ osd_hosts }}"
       delegate_to: "{{ item.1 }}"
 
     - name: set osd(s) out when ceph-disk deactivating fail
       command: ceph --cluster {{ cluster }} osd out osd.{{ item.0 }}
+      delegate_to: "{{ groups[mon_group_name][0] }}"
       with_together:
-      - "{{ osd_ids.split(',') }}"
+        - "{{ osd_to_kill.split(',') }}"
         - "{{ deactivate.results }}"
       when:
         - item.1.stderr|length > 0
@@ -138,30 +115,41 @@
       register: destroy
       ignore_errors: yes
       with_together:
-      - "{{ osd_ids.split(',') }}"
+        - "{{ osd_to_kill.split(',') }}"
-      - "{{ real_ips }}"
+        - "{{ osd_hosts }}"
       delegate_to: "{{ item.1 }}"
 
     - name: remove osd(s) from crush_map when ceph-disk destroy fail
       command: ceph --cluster {{ cluster }} osd crush remove osd.{{ item.0 }}
+      delegate_to: "{{ groups[mon_group_name][0] }}"
       with_together:
-      - "{{ osd_ids.split(',') }}"
+        - "{{ osd_to_kill.split(',') }}"
         - "{{ destroy.results }}"
       when:
         - item.1.stderr|length > 0
 
     - name: delete osd(s) auth key when ceph-disk destroy fail
       command: ceph --cluster {{ cluster }} auth del osd.{{ item.0 }}
+      delegate_to: "{{ groups[mon_group_name][0] }}"
       with_together:
-      - "{{ osd_ids.split(',') }}"
+        - "{{ osd_to_kill.split(',') }}"
         - "{{ destroy.results }}"
       when:
         - item.1.stderr|length > 0
 
     - name: deallocate osd(s) id when ceph-disk destroy fail
       command: ceph --cluster {{ cluster }} osd rm {{ item.0 }}
+      delegate_to: "{{ groups[mon_group_name][0] }}"
       with_together:
-      - "{{ osd_ids.split(',') }}"
+        - "{{ osd_to_kill.split(',') }}"
         - "{{ destroy.results }}"
       when:
         - item.1.stderr|length > 0
+
+    - name: show ceph health
+      command: ceph --cluster {{ cluster }} -s
+      delegate_to: "{{ groups[mon_group_name][0] }}"
+
+    - name: show ceph osd tree
+      command: ceph --cluster {{ cluster }} osd tree
+      delegate_to: "{{ groups[mon_group_name][0] }}"

tests/functional/centos/7/cluster/group_vars/osds

@@ -1,4 +1,5 @@
 ---
+copy_admin_key: true
 os_tuning_params:
   - { name: kernel.pid_max, value: 4194303 }
   - { name: fs.file-max, value: 26234859 }

tests/functional/centos/7/cluster/hosts

@@ -1,19 +1,19 @@
 [mons]
-mon0 monitor_address=192.168.1.10
+ceph-mon0 monitor_address=192.168.1.10
-mon1 monitor_interface=eth1
+ceph-mon1 monitor_interface=eth1
-mon2 monitor_address=192.168.1.12
+ceph-mon2 monitor_address=192.168.1.12
 
 [osds]
-osd0
+ceph-osd0
 
 [mdss]
-mds0
+ceph-mds0
 
 [rgws]
-rgw0
+ceph-rgw0
 
 [clients]
-client0
+ceph-client0
 
 [mgrs]
-mgr0
+ceph-mgr0

tests/functional/centos/7/cluster/vagrant_variables.yml

@@ -72,3 +72,6 @@ vagrant_disable_synced_folder: true
 os_tuning_params:
   - { name: kernel.pid_max, value: 4194303 }
   - { name: fs.file-max, value: 26234859 }
+
+# VM prefix name, need to match the hostname
+label_prefix: ceph

tests/functional/tests/mon/test_mons.py

@@ -11,13 +11,13 @@ class TestMons(object):
         assert host.socket("tcp://%s:6789" % node["address"]).is_listening
 
     def test_mon_service_is_running(self, node, host):
-        service_name = "ceph-mon@ceph-{hostname}".format(
+        service_name = "ceph-mon@{hostname}".format(
             hostname=node["vars"]["inventory_hostname"]
         )
         assert host.service(service_name).is_running
 
     def test_mon_service_is_enabled(self, node, host):
-        service_name = "ceph-mon@ceph-{hostname}".format(
+        service_name = "ceph-mon@{hostname}".format(
             hostname=node["vars"]["inventory_hostname"]
         )
         assert host.service(service_name).is_enabled
@@ -32,7 +32,7 @@ class TestMons(object):
         assert File(node["conf_path"]).contains("^mon initial members = .*$")
 
     def test_initial_members_line_has_correct_value(self, node, File):
-        mons = ",".join("ceph-%s" % host
+        mons = ",".join("%s" % host
                         for host in node["vars"]["groups"]["mons"])
         line = "mon initial members = {}".format(mons)
         assert File(node["conf_path"]).contains(line)
@@ -49,7 +49,7 @@ class TestOSDs(object):
 
     @pytest.mark.docker
     def test_all_docker_osds_are_up_and_in(self, node, host):
-        cmd = "sudo docker exec ceph-mon-ceph-{} ceph --cluster={} --connect-timeout 5 -s".format(
+        cmd = "sudo docker exec ceph-mon-{} ceph --cluster={} --connect-timeout 5 -s".format(
             node["vars"]["inventory_hostname"],
             node["cluster_name"]
         )

tox.ini

@@ -1,6 +1,6 @@
 [tox]
 envlist = {dev,jewel,luminous,rhcs}-{ansible2.2,ansible2.3}-{xenial_cluster,journal_collocation,centos7_cluster,dmcrypt_journal,dmcrypt_journal_collocation,docker_cluster,purge_cluster,purge_dmcrypt,docker_dedicated_journal,docker_dmcrypt_journal_collocation,update_dmcrypt,update_cluster,cluster,purge_docker_cluster,update_docker_cluster}
-  {dev,luminous}-{ansible2.2,ansible2.3}-{bluestore_journal_collocation,bluestore_cluster,bluestore_dmcrypt_journal,bluestore_dmcrypt_journal_collocation,bluestore_docker_cluster,bluestore_docker_dedicated_journal,bluestore_docker_dmcrypt_journal_collocation,lvm_osds,purge_lvm_osds}
+  {dev,luminous}-{ansible2.2,ansible2.3}-{bluestore_journal_collocation,bluestore_cluster,bluestore_dmcrypt_journal,bluestore_dmcrypt_journal_collocation,bluestore_docker_cluster,bluestore_docker_dedicated_journal,bluestore_docker_dmcrypt_journal_collocation,lvm_osds,purge_lvm_osds,shrink_mon,shrink_osd}
 
 skipsdist = True
 
@@ -81,6 +81,21 @@ commands=
 
   testinfra -n 4 --sudo -v --connection=ansible --ansible-inventory={changedir}/hosts {toxinidir}/tests/functional/tests
 
+[shrink-mon]
+commands=
+  cp {toxinidir}/infrastructure-playbooks/shrink-mon.yml {toxinidir}/shrink-mon.yml
+  ansible-playbook -vv -i {changedir}/hosts {toxinidir}/shrink-mon.yml --extra-vars "\
+      ireallymeanit=yes \
+      mon_to_kill=ceph-mon2 \
+  "
+[shrink-osd]
+commands=
+  cp {toxinidir}/infrastructure-playbooks/shrink-osd.yml {toxinidir}/shrink-osd.yml
+  ansible-playbook -vv -i {changedir}/hosts {toxinidir}/shrink-osd.yml --extra-vars "\
+      ireallymeanit=yes \
+      osd_to_kill=0 \
+  "
+
 [testenv]
 whitelist_externals =
     vagrant
@@ -128,6 +143,8 @@ changedir=
   dmcrypt_journal_collocation: {toxinidir}/tests/functional/centos/7/crypt-jrn-col
   # tests a 1 mon, 1 osd, 1 mds and 1 rgw centos7 cluster using non-collocated OSD scenario
   centos7_cluster: {toxinidir}/tests/functional/centos/7/cluster
+  shrink_mon: {toxinidir}/tests/functional/centos/7/cluster
+  shrink_osd: {toxinidir}/tests/functional/centos/7/cluster
   # an alias for centos7_cluster, this makes the name better suited for rhcs testing
   cluster: {toxinidir}/tests/functional/centos/7/cluster
   # tests a 1 mon, 1 osd, 1 mds and 1 rgw centos7 cluster using docker
@@ -183,5 +200,7 @@ commands=
   update_dmcrypt: {[update]commands}
   update_cluster: {[update]commands}
   update_docker_cluster: {[update]commands}
+  shrink_mon: {[shrink-mon]commands}
+  shrink_osd: {[shrink-osd]commands}
 
   vagrant destroy --force