wip

Signed-off-by: Seena Fallah <[email protected]>
ceph · Feb 18, 2024 · ffe9b2f · ffe9b2f
1 parent 079c33c
commit ffe9b2f
Show file tree

Hide file tree

Showing 3 changed files with 38 additions and 19 deletions.
diff --git a/roles/ceph-crash/templates/ceph-crash.service.j2 b/roles/ceph-crash/templates/ceph-crash.service.j2
@@ -24,9 +24,9 @@ ExecStart=/usr/bin/{{ container_binary }} run --rm --name ceph-crash-%i \
 {% if cluster != 'ceph' %}
 -e CEPH_ARGS="--cluster {{ cluster }}" \
 {% endif %}
-{% for v in ceph_common_container_params['volumes'] %}
-  -v {{ v }} \
-{% endfor %}
+-v /var/lib/ceph/crash:/var/lib/ceph/crash:z \
+-v /etc/localtime:/etc/localtime:ro \
+-v /etc/ceph:/etc/ceph:z \
 {% for k, v in ceph_common_container_params['envs'].items() %}
   -e {{ k }}={{ v }} \
 {% endfor %}

diff --git a/roles/ceph-mon/tasks/ceph_keys.yml b/roles/ceph-mon/tasks/ceph_keys.yml
@@ -1,20 +1,39 @@
 ---
 - name: Waiting for the monitor(s) to form the quorum...
-  ansible.builtin.command: >
-    {{ container_exec_cmd }}
-    ceph
-    --cluster {{ cluster }}
-    daemon mon.{{ ansible_facts['hostname'] }}
-    mon_status
-    --format json
-  register: ceph_health_raw
-  run_once: true
-  until: >
-    (ceph_health_raw.stdout | length > 0) and (ceph_health_raw.stdout | default('{}') | from_json)['state'] in ['leader', 'peon']
-  retries: "{{ handler_health_mon_check_retries }}"
-  delay: "{{ handler_health_mon_check_delay }}"
-  changed_when: false
-  when: not ansible_check_mode
+  block:
+    - name: Wait for the monitor(s) to form the quorum
+      ansible.builtin.command: >
+        {{ container_exec_cmd }}
+        ceph
+        --cluster {{ cluster }}
+        daemon mon.{{ ansible_facts['hostname'] }}
+        mon_status
+        --format json
+      register: ceph_health_raw
+      run_once: true
+      until: >
+        (ceph_health_raw.stdout | length > 0) and (ceph_health_raw.stdout | default('{}') | from_json)['state'] in ['leader', 'peon']
+      retries: "{{ handler_health_mon_check_retries }}"
+      delay: "{{ handler_health_mon_check_delay }}"
+      changed_when: false
+      when: not ansible_check_mode
+  rescue:
+    - name: List run dir
+      ansible.builtin.command: ls -lah /var/run/ceph
+      changed_when: false
+    - name: List run dir inside container
+      ansible.builtin.command: "{{ container_exec_cmd }} ls -lah /var/run/ceph"
+      changed_when: false
+    - name: Ceph status
+      ansible.builtin.command: >
+        {{ container_exec_cmd }}
+        ceph
+        --cluster {{ cluster }}
+        -s
+      changed_when: false
+    - name: Fail the task
+      ansible.builtin.fail:
+        msg: "Monitor(s) failed to form the quorum"
 
 - name: Fetch ceph initial keys
   ceph_key:

diff --git a/tox.ini b/tox.ini
@@ -360,7 +360,7 @@ commands=
   all_daemons,all_daemons_ipv6,collocation: py.test --reruns 20 --reruns-delay 3 -n 8 --durations=0 --sudo -v --connection=ansible --ansible-inventory={changedir}/{env:INVENTORY} --ssh-config={changedir}/vagrant_ssh_config {toxinidir}/tests/functional/tests
 
   # handlers/idempotency test
-  all_daemons,all_daemon_ipv6,all_in_one,collocation: ansible-playbook -vv --diff -i {changedir}/{env:INVENTORY} {toxinidir}/{env:PLAYBOOK:site.yml.sample} --extra-vars "delegate_facts_host={env:DELEGATE_FACTS_HOST:True} ceph_docker_image_tag={env:CEPH_DOCKER_IMAGE_TAG_BIS:latest-bis-main} ceph_dev_branch={env:CEPH_DEV_BRANCH:main} ceph_dev_sha1={env:CEPH_DEV_SHA1:latest} yes_i_know=true" --extra-vars @ceph-override.json
+  all_daemons,all_daemon_ipv6,all_in_one,collocation: ansible-playbook -vv --diff -i {changedir}/{env:INVENTORY} {toxinidir}/{env:PLAYBOOK:site.yml.sample} --extra-vars "no_log_on_ceph_key_tasks=false delegate_facts_host={env:DELEGATE_FACTS_HOST:True} ceph_docker_image_tag={env:CEPH_DOCKER_IMAGE_TAG_BIS:latest-bis-main} ceph_dev_branch={env:CEPH_DEV_BRANCH:main} ceph_dev_sha1={env:CEPH_DEV_SHA1:latest} yes_i_know=true" --extra-vars @ceph-override.json
 
   purge: {[purge]commands}
   purge_dashboard: {[purge-dashboard]commands}