deep_scrub_choked

2026-04-06 07:08:17 +03:00
commit 9149dc1d7f
840 changed files with 52471 additions and 0 deletions
@@ -0,0 +1,14 @@
+---
+galaxy_info:
+  company: Red Hat
+  author: Guillaume Abrioux
+  description: Deploy ceph-crash
+  license: Apache
+  min_ansible_version: '2.7'
+  platforms:
+    - name: EL
+      versions:
+        - 'all'
+  galaxy_tags:
+    - system
+dependencies: []
@@ -0,0 +1,67 @@
+---
+- name: Create and copy client.crash keyring
+  when: cephx | bool
+  block:
+    - name: Create client.crash keyring
+      ceph_key:
+        name: "client.crash"
+        caps:
+          mon: 'allow profile crash'
+          mgr: 'allow profile crash'
+        cluster: "{{ cluster }}"
+        dest: "{{ ceph_conf_key_directory }}"
+        import_key: true
+        mode: "{{ ceph_keyring_permissions }}"
+        owner: "{{ ceph_uid if containerized_deployment | bool else 'ceph' }}"
+        group: "{{ ceph_uid if containerized_deployment | bool else 'ceph' }}"
+      environment:
+        CEPH_CONTAINER_IMAGE: "{{ ceph_docker_registry + '/' + ceph_docker_image + ':' + ceph_docker_image_tag if containerized_deployment | bool else None }}"
+        CEPH_CONTAINER_BINARY: "{{ container_binary }}"
+      delegate_to: "{{ groups.get(mon_group_name, [])[0] }}"
+      run_once: true
+      no_log: "{{ no_log_on_ceph_key_tasks }}"
+
+    - name: Get keys from monitors
+      ceph_key_info:
+        name: client.crash
+        cluster: "{{ cluster }}"
+        output_format: plain
+        state: info
+      environment:
+        CEPH_CONTAINER_IMAGE: "{{ ceph_docker_registry + '/' + ceph_docker_image + ':' + ceph_docker_image_tag if containerized_deployment | bool else None }}"
+        CEPH_CONTAINER_BINARY: "{{ container_binary }}"
+      register: _crash_keys
+      delegate_to: "{{ groups.get(mon_group_name)[0] }}"
+      run_once: true
+      no_log: "{{ no_log_on_ceph_key_tasks }}"
+
+    - name: Copy ceph key(s) if needed
+      ansible.builtin.copy:
+        dest: "{{ ceph_conf_key_directory }}/{{ cluster }}.client.crash.keyring"
+        content: "{{ _crash_keys.stdout + '\n' }}"
+        owner: "{{ ceph_uid if containerized_deployment | bool else 'ceph' }}"
+        group: "{{ ceph_uid if containerized_deployment | bool else 'ceph' }}"
+        mode: "{{ ceph_keyring_permissions }}"
+      no_log: "{{ no_log_on_ceph_key_tasks }}"
+
+- name: Start ceph-crash daemon
+  when: containerized_deployment | bool
+  block:
+    - name: Create /var/lib/ceph/crash/posted
+      ansible.builtin.file:
+        path: /var/lib/ceph/crash/posted
+        state: directory
+        mode: '0755'
+        owner: "{{ ceph_uid }}"
+        group: "{{ ceph_uid }}"
+
+    - name: Include_tasks systemd.yml
+      ansible.builtin.include_tasks: systemd.yml
+
+- name: Start the ceph-crash service
+  ansible.builtin.systemd:
+    name: "{{ 'ceph-crash@' + ansible_facts['hostname'] if containerized_deployment | bool else 'ceph-crash.service' }}"
+    state: started
+    enabled: true
+    masked: false
+    daemon_reload: true
@@ -0,0 +1,9 @@
+---
+- name: Generate systemd unit file for ceph-crash container
+  ansible.builtin.template:
+    src: "{{ role_path }}/templates/ceph-crash.service.j2"
+    dest: /etc/systemd/system/ceph-crash@.service
+    owner: "root"
+    group: "root"
+    mode: "0644"
+  notify: Restart ceph crash
@@ -0,0 +1,52 @@
+[Unit]
+Description=Ceph crash dump collector
+{% if container_binary == 'docker' %}
+After=docker.service network-online.target local-fs.target time-sync.target
+Requires=docker.service
+{% else %}
+After=network-online.target local-fs.target time-sync.target
+{% endif %}
+Wants=network-online.target local-fs.target time-sync.target
+
+[Service]
+{% if container_binary == 'podman' %}
+ExecStartPre=-/usr/bin/rm -f /%t/%n-pid /%t/%n-cid
+ExecStartPre=-/usr/bin/{{ container_binary }} rm --storage ceph-crash-%i
+{% endif %}
+ExecStartPre=-/usr/bin/{{ container_binary }} rm -f ceph-crash-%i
+ExecStart=/usr/bin/{{ container_binary }} run --rm --name ceph-crash-%i \
+{% if container_binary == 'podman' %}
+-d --log-driver journald --conmon-pidfile /%t/%n-pid --cidfile /%t/%n-cid \
+{% endif %}
+--pids-limit={{ 0 if container_binary == 'podman' else -1 }} \
+--security-opt label=disable \
+--net=host \
+{% if cluster != 'ceph' %}
+-e CEPH_ARGS="--cluster {{ cluster }}" \
+{% endif %}
+{% for v in ceph_common_container_params['volumes'] %}
+  -v {{ v }} \
+{% endfor %}
+{% for k, v in ceph_common_container_params['envs'].items() %}
+  -e {{ k }}={{ v }} \
+{% endfor %}
+--entrypoint=/usr/bin/ceph-crash {{ ceph_docker_registry }}/{{ ceph_docker_image }}:{{ ceph_docker_image_tag }}
+{% if container_binary == 'podman' %}
+ExecStop=-/usr/bin/sh -c "/usr/bin/{{ container_binary }} rm -f `cat /%t/%n-cid`"
+{% else %}
+ExecStop=-/usr/bin/{{ container_binary }} stop ceph-crash-%i
+{% endif %}
+StartLimitInterval=10min
+StartLimitBurst=30
+{% if container_binary == 'podman' %}
+Type=forking
+PIDFile=/%t/%n-pid
+{% endif %}
+KillMode=none
+Restart=always
+RestartSec=10s
+TimeoutStartSec=120
+TimeoutStopSec=10
+
+[Install]
+WantedBy=ceph.target