dashboard: align the way containers are managed

This commit aligns the way the different containers are managed with how
it's currently done with the other ceph daemon.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit cc285c417a)

roles/ceph-grafana/handlers/main.yml

@@ -1,8 +0,0 @@
----
-- name: enable service
-  # We use the systemd module here so we can use the daemon_reload feature,
-  # since we're shipping the .service file ourselves
-  systemd:
-    name: grafana-server
-    daemon_reload: true
-    enabled: true

roles/ceph-grafana/tasks/setup_container.yml

@@ -16,31 +16,6 @@
     - /etc/grafana
     - /var/lib/grafana
 
-- name: make sure the grafana-server service is down
-  service:
-    name: grafana-server
-    state: stopped
-  failed_when: false
-
-# Make sure we re-create the container
-- name: remove old grafana-server container
-  command: "{{ container_binary }} rm -f grafana-server"
-  changed_when: false
-  failed_when: false
-
-- name: create grafana-server container
-  shell: |
-    {{ container_binary }} create --name grafana-server \
-      -v "/etc/grafana:/etc/grafana:Z" \
-      -v "/var/lib/grafana:/var/lib/grafana:Z" \
-      "--net=host" \
-      "--cpu-period={{ grafana_container_cpu_period }}" \
-      "--cpu-quota={{ grafana_container_cpu_period * grafana_container_cpu_cores }}" \
-      "--memory={{ grafana_container_memory }}GB" \
-      "--memory-swap={{ grafana_container_memory * 2 }}GB" \
-      -e "GF_INSTALL_PLUGINS={{ grafana_plugins|join(',') }}" \
-      "{{ grafana_container_image }}"
-
 - name: ship systemd service
   template:
     src: grafana-server.service.j2
@@ -48,4 +23,11 @@
     owner: root
     group: root
     mode: 0644
-  notify: enable service
+
+- name: start the grafana-server service
+  systemd:
+    name: grafana-server
+    state: started
+    enabled: yes
+    daemon_reload: yes
+  failed_when: false

roles/ceph-grafana/templates/grafana-server.service.j2

@@ -6,14 +6,25 @@ Description=grafana-server
 After=docker.service
 {% endif %}
 
- [Service]	
+[Service]
 EnvironmentFile=-/etc/environment
-ExecStart=/usr/bin/{{ container_binary }} start --attach grafana-server	
+ExecStartPre=-/usr/bin/{{ container_binary }} stop grafana-server
+ExecStartPre=-/usr/bin/{{ container_binary }} rm grafana-server
+ExecStart=/usr/bin/{{ container_binary }} run --rm --name=grafana-server \
+  -v /etc/grafana:/etc/grafana:Z \
+  -v /var/lib/grafana:/var/lib/grafana:Z \
+  --net=host \
+  --cpu-period={{ grafana_container_cpu_period }} \
+  --cpu-quota={{ grafana_container_cpu_period * grafana_container_cpu_cores }} \
+  --memory={{ grafana_container_memory }}GB \
+  --memory-swap={{ grafana_container_memory * 2 }}GB \
+  -e GF_INSTALL_PLUGINS={{ grafana_plugins|join(',') }} \
+  {{ grafana_container_image }}
 ExecStop=-/usr/bin/{{ container_binary }} stop grafana-server
 Restart=always
 RestartSec=10s
 TimeoutStartSec=120
 TimeoutStopSec=15
 
- [Install]	
+[Install]
 WantedBy=multi-user.target

roles/ceph-handler/handlers/main.yml

@@ -458,13 +458,3 @@
       set_fact:
          _rbd_target_api_handler_called: False
       listen: "restart ceph rbd-target-api"
-
-    - name: restart node-exporter service
-      listen: "restart node-exporter service"
-      # We use the systemd module here so we can use the daemon_reload feature,
-      # since we're shipping the .service file ourselves
-      systemd:
-        name: 'node_exporter'
-        daemon_reload: true
-        enabled: true
-        state: restarted

roles/ceph-node-exporter/tasks/setup_container.yml

@@ -1,27 +1,4 @@
 ---
-- name: make sure the node_exporter service is down
-  service:
-    name: node_exporter
-    state: stopped
-  failed_when: false
-
-# Make sure we re-create the container
-- name: remove old node-exporter container
-  command: "{{ container_binary }} rm -f node-exporter"
-  changed_when: false
-  failed_when: false
-
-- name: start node-exporter container
-  shell: |
-    {{ container_binary }} run --detach  --name node-exporter \
-      -v /proc:/host/proc:ro -v /sys:/host/sys:ro \
-      --net=host \
-      {{ node_exporter_container_image }} \
-      '--path.procfs=/host/proc' \
-      '--path.sysfs=/host/sys' \
-      '--no-collector.timex'
-  notify: restart node-exporter service
-
 - name: ship systemd service
   template:
     src: node_exporter.service.j2
@@ -29,4 +6,11 @@
     owner: root
     group: root
     mode: 0644
-  notify: restart node-exporter service
+
+- name: start the node_exporter service
+  systemd:
+    name: node_exporter
+    state: started
+    enabled: yes
+    daemon_reload: yes
+  failed_when: false

roles/ceph-node-exporter/templates/node_exporter.service.j2

@@ -8,7 +8,14 @@ After=docker.service
 
 [Service]
 EnvironmentFile=-/etc/environment
-ExecStart=/usr/bin/{{ container_binary }} start --attach node-exporter
+ExecStartPre=-/usr/bin/{{ container_binary }} rm -f node-exporter
+ExecStart=/usr/bin/{{ container_binary }} run --name=node-exporter \
+  -v /proc:/host/proc:ro -v /sys:/host/sys:ro \
+  --net=host \
+  --path.procfs=/host/proc \
+  --path.sysfs=/host/sys \
+  --no-collector.timex \
+    {{ node_exporter_container_image }}
 # Make sure the cfg80211 is loaded before running the container, the node
 # exporter needs this module loaded to test for presence of wi-fi devices
 ExecStartPre=/usr/sbin/modprobe cfg80211

roles/ceph-prometheus/tasks/setup_container.yml

@@ -1,60 +1,4 @@
 ---
-- name: make sure the alertmanager service is down
-  service:
-    name: alertmanager
-    state: stopped
-  failed_when: false
-
-# Make sure we re-create the container
-- name: remove old alertmanager container
-  command: "{{ container_binary }} rm -f alertmanager"
-  changed_when: false
-  failed_when: false
-
-- name: start alertmanager container
-  shell: |
-    {{ container_binary }} run --detach  --name alertmanager \
-      -v "{{ alertmanager_conf_dir }}:/etc/alertmanager:Z" \
-      -v "{{ alertmanager_data_dir }}:/alertmanager:Z" \
-      "--net=host" \
-      "--cpu-period={{ alertmanager_container_cpu_period }}" \
-      "--cpu-quota={{ alertmanager_container_cpu_period * alertmanager_container_cpu_cores }}" \
-      "--memory={{ alertmanager_container_memory }}GB" \
-      "--memory-swap={{ alertmanager_container_memory * 2 }}GB" \
-      "{{ alertmanager_container_image }}" \
-      "--config.file=/etc/alertmanager/alertmanager.yml" \
-      "--storage.path=/alertmanager"
-  notify: service handler
-
-- name: make sure the prometheus service is down
-  service:
-    name: prometheus
-    state: stopped
-  failed_when: false
-
-# Make sure we re-create the container
-- name: remove old prometheus container
-  command: "{{ container_binary }} rm -f prometheus"
-  changed_when: false
-  failed_when: false
-
-- name: start prometheus container
-  shell: |
-    {{ container_binary }} run --detach  --name prometheus \
-      -v "{{ prometheus_conf_dir }}:/etc/prometheus:Z" \
-      -v "{{ prometheus_data_dir }}:/prometheus:Z" \
-      "--net=host" \
-      "--user={{ prometheus_user_id }}" \
-      "--cpu-period={{ prometheus_container_cpu_period }}" \
-      "--cpu-quota={{ prometheus_container_cpu_period * prometheus_container_cpu_cores }}" \
-      "--memory={{ prometheus_container_memory }}GB" \
-      "--memory-swap={{ prometheus_container_memory * 2 }}GB" \
-      "{{ prometheus_container_image }}" \
-      "--config.file=/etc/prometheus/prometheus.yml" \
-      "--storage.tsdb.path=/prometheus" \
-      "--web.external-url=http://{{ inventory_hostname }}:9090/"
-  notify: service handler
-
 - name: ship systemd services
   template:
     src: "{{ item }}.j2"
@@ -66,3 +10,13 @@
     - 'alertmanager.service'
     - 'prometheus.service'
   notify: service handler
+
+- name: start prometheus services
+  systemd:
+    name: "{{ item }}"
+    daemon_reload: true
+    enabled: true
+    state: started
+  with_items:
+    - prometheus
+    - alertmanager

roles/ceph-prometheus/templates/alertmanager.service.j2

@@ -8,7 +8,18 @@ After=docker.service
 
 [Service]
 EnvironmentFile=-/etc/environment
-ExecStart=/usr/bin/{{ container_binary }} start --attach alertmanager
+ExecStartPre=-/usr/bin/{{ container_binary }} rm -f alertmanager
+ExecStart=/usr/bin/{{ container_binary }} run --name=alertmanager \
+  -v "{{ alertmanager_conf_dir }}:/etc/alertmanager:Z" \
+  -v "{{ alertmanager_data_dir }}:/alertmanager:Z" \
+  --net=host \
+  --cpu-period={{ alertmanager_container_cpu_period }} \
+  --cpu-quota={{ alertmanager_container_cpu_period * alertmanager_container_cpu_cores }} \
+  --memory={{ alertmanager_container_memory }}GB \
+  --memory-swap={{ alertmanager_container_memory * 2 }}GB \
+  {{ alertmanager_container_image }} \
+  --config.file=/etc/alertmanager/alertmanager.yml \
+  --storage.path=/alertmanager"
 ExecStop=/usr/bin/{{ container_binary }} stop alertmanager
 Restart=always
 RestartSec=10s

roles/ceph-prometheus/templates/prometheus.service.j2

@@ -8,7 +8,20 @@ After=docker.service
 
 [Service]
 EnvironmentFile=-/etc/environment
-ExecStart=/usr/bin/{{ container_binary }} start --attach prometheus
+ExecStartPre=-/usr/bin/{{ container_binary }} rm -f prometheus
+ExecStart=/usr/bin/{{ container_binary }} run --name=prometheus \
+  -v "{{ prometheus_conf_dir }}:/etc/prometheus:Z" \
+  -v "{{ prometheus_data_dir }}:/prometheus:Z" \
+  --net=host \
+  --user={{ prometheus_user_id }} \
+  --cpu-period={{ prometheus_container_cpu_period }} \
+  --cpu-quota={{ prometheus_container_cpu_period * prometheus_container_cpu_cores }} \
+  --memory={{ prometheus_container_memory }}GB \
+  --memory-swap={{ prometheus_container_memory * 2 }}GB \
+  {{ prometheus_container_image }} \
+  --config.file=/etc/prometheus/prometheus.yml \
+  --storage.tsdb.path=/prometheus \
+  --web.external-url=http://{{ inventory_hostname }}:9090/"
 ExecStop=/usr/bin/{{ container_binary }} stop prometheus
 Restart=always
 RestartSec=10s