Merge pull request #1178 from zhsj/dev-partition

Add prepare osd with partition devices in raw_multi_journal

infrastructure-playbooks/make-osd-partitions.yml

@@ -0,0 +1,49 @@
+---
+# This playbook will make custom partition layout for your osd hosts.
+# You should define `devices` variable for every host.
+#
+# For example, in host_vars/hostname1
+#
+# devices:
+# - device_name: sdb
+#   partitions:
+#   - index: 1
+#     size: 10G
+#     type: data
+#   - index: 2
+#     size: 5G
+#     type: journal
+# - device_name: sdc
+#   partitions:
+#   - index: 1
+#     size: 10G
+#     type: data
+#   - index: 2
+#     size: 5G
+#     type: journal
+#
+- vars:
+    osd_group_name: osds
+    journal_typecode: 45b0969e-9b03-4f30-b4c6-b4b80ceff106
+    data_typecode: 4fbd7e29-9d25-41b8-afd0-062c0ceff05d
+  hosts:
+  - "{{ osd_group_name }}"
+
+  tasks:
+  - name: install sgdisk(gdisk)
+    package:
+      name: gdisk
+      state: present
+
+  - name: erase all previous partitions(dangerous!!!)
+    shell: sgdisk --zap-all -- /dev/{{item.device_name}}
+    with_items: "{{ devices }}"
+
+  - name: make osd partitions
+    shell: >
+           sgdisk --new={{item.1.index}}:0:+{{item.1.size}} "--change-name={{item.1.index}}:ceph {{item.1.type}}"
+           "--typecode={{item.1.index}}:{% if item.1.type=='data' %}{{data_typecode}}{% else %}{{journal_typecode}}{% endif %}"
+           --mbrtogpt -- /dev/{{item.0.device_name}}
+    with_subelements:
+    - "{{ devices }}"
+    - partitions

roles/ceph-osd/tasks/check_devices.yml

@@ -24,14 +24,26 @@
   # Hard code this so we will skip the entire file instead of individual tasks (Default isn't Consistent)
   static: False
 
+- name: check the journal device is partition
+  shell: "readlink -f {{ item }} | egrep '/dev/([hsv]d[a-z]{1,2}|cciss/c[0-9]d[0-9]p|nvme[0-9]n[0-9]p)[0-9]{1,2}$'"
+  with_items: "{{ raw_journal_devices }}"
+  changed_when: false
+  failed_when: false
+  always_run: true
+  register: journal_ispartition_results
+
 - name: check the partition status of the journal devices
-  shell: "parted --script {{ item }} print > /dev/null 2>&1"
-  with_items: "{{ raw_journal_devices|unique }}"
+  shell: "parted --script {{ item.1 }} print > /dev/null 2>&1"
+  with_together:
+    - "{{ journal_ispartition_results.results }}"
+    - "{{ raw_journal_devices|unique }}"
   changed_when: false
   failed_when: false
   always_run: true
   register: journal_partition_status
-  when: raw_multi_journal or dmcrypt_dedicated_journal
+  when:
+    - (raw_multi_journal or dmcrypt_dedicated_journal)
+    - item.0.rc != 0
 
 - name: fix partitions gpt header or labels of the journal devices
   shell: "sgdisk --zap-all --clear --mbrtogpt -- {{ item.1 }} || sgdisk --zap-all --clear --mbrtogpt -- {{ item.1 }}"
@@ -40,5 +52,6 @@
     - "{{ raw_journal_devices|unique }}"
   changed_when: false
   when:
-    - raw_multi_journal or dmcrypt_dedicated_journal
-    - item.0.rc != 0
+    - (raw_multi_journal or dmcrypt_dedicated_journal)
+    - not item.0.get("skipped")
+    - item.0.get("rc", 0) != 0

roles/ceph-osd/tasks/check_devices_static.yml

@@ -8,12 +8,15 @@
   register: ispartition_results
 
 - name: check the partition status of the osd disks
-  shell: "parted --script {{ item }} print > /dev/null 2>&1"
-  with_items: "{{ devices }}"
+  shell: "parted --script {{ item.1 }} print > /dev/null 2>&1"
+  with_together:
+    - "{{ ispartition_results.results }}"
+    - "{{ devices }}"
   changed_when: false
   failed_when: false
   always_run: true
   register: osd_partition_status_results
+  when: item.0.rc != 0
 
 # NOTE: The following calls to sgdisk are retried because sgdisk is known to
 # fully wipe a device the first time around. There is no need to halt execution
@@ -26,12 +29,17 @@
     - "{{ osd_partition_status_results.results }}"
     - "{{ devices }}"
   changed_when: false
-  when: item.0.rc != 0
+  when:
+    - not item.0.get("skipped")
+    - item.0.get("rc", 0) != 0
 
 - name: check if a partition named 'ceph' exists
-  shell: "parted --script {{ item }} print | egrep -sq '^ 1.*ceph'"
-  with_items: "{{ devices }}"
+  shell: "parted --script {{ item.1 }} print | egrep -sq '^ 1.*ceph'"
+  with_together:
+    - "{{ ispartition_results.results }}"
+    - "{{ devices }}"
   changed_when: false
   failed_when: false
   always_run: true
   register: parted_results
+  when: item.0.rc != 0

roles/ceph-osd/tasks/scenarios/raw_multi_journal.yml

@@ -7,21 +7,15 @@
 # if you have 64 disks with 4TB each, this will take a while
 # since Ansible will sequential process the loop
 
-# NOTE (alahouze): if the device is a partition, the parted command below has
-# failed, this is why we check if the device is a partition too.
 - name: prepare osd disk(s)
-  command: "ceph-disk prepare --cluster {{ cluster }} {{ item.2 }} {{ item.3 }}"
+  command: "ceph-disk prepare --cluster {{ cluster }} {{ item.1 }} {{ item.2 }}"
   with_together:
     - "{{ parted_results.results }}"
-    - "{{ ispartition_results.results }}"
     - "{{ devices }}"
     - "{{ raw_journal_devices }}"
   changed_when: false
   when:
-    - not item.0.get("skipped")
-    - not item.1.get("skipped")
-    - item.0.get("rc", 0) != 0
-    - item.1.get("rc", 0) != 0
+    - item.0.get("skipped") or item.0.get("rc", 0) != 0
     - raw_multi_journal
     - not osd_auto_discovery