Merge pull request #398 from mattymo/rework_systemd_wait_master

Improve systemd handling and stabilize docker

roles/docker/handlers/main.yml

@@ -4,6 +4,8 @@
   notify:
     - Docker | reload systemd
     - Docker | reload docker
+    - Docker | pause while Docker restarts
+    - Docker | wait for docker
 
 - name : Docker | reload systemd
   shell: systemctl daemon-reload
@@ -13,3 +15,13 @@
   service:
     name: docker
     state: restarted
+
+- name: Docker | pause while Docker restarts
+  pause: seconds=10 prompt="Waiting for docker restart"
+
+- name: Docker | wait for docker
+  command: /usr/bin/docker images
+  register: docker_ready
+  retries: 10
+  delay: 5
+  until: docker_ready.rc == 0

roles/docker/tasks/main.yml

@@ -59,6 +59,14 @@
   when: ansible_service_mgr == "systemd" and
         (http_proxy is defined or https_proxy is defined or no_proxy is defined)
 
+- name: Write docker.service systemd file
+  template:
+    src: systemd-docker.service.j2
+    dest: /etc/systemd/system/docker.service
+  register: docker_service_file
+  notify: restart docker
+  when: ansible_service_mgr == "systemd" and ansible_os_family != "CoreOS"
+
 - meta: flush_handlers
 
 - name: ensure docker service is started and enabled

roles/docker/templates/systemd-docker.service.j2

@@ -2,11 +2,11 @@
 Description=Docker Application Container Engine
 Documentation=http://docs.docker.com
 {% if ansible_os_family == "RedHat" %}
-After=network.target
+After=network.target docker-storage-setup.service
 Wants=docker-storage-setup.service
 {% elif ansible_os_family == "Debian" %}
 After=network.target docker.socket
-Requires=docker.socket
+Wants=docker.socket
 {% endif %}
 
 [Service]
@@ -20,17 +20,21 @@ EnvironmentFile=-/etc/sysconfig/docker-storage
 EnvironmentFile=-/etc/default/docker
 {% endif %}
 Environment=GOTRACEBACK=crash
+ExecReload=/bin/kill -s HUP $MAINPID
+Delegate=yes
+KillMode=process
 ExecStart=/usr/bin/docker daemon \
           $OPTIONS \
           $DOCKER_STORAGE_OPTIONS \
           $DOCKER_NETWORK_OPTIONS \
           $INSECURE_REGISTRY \
           $DOCKER_OPTS
+TasksMax=infinity
 LimitNOFILE=1048576
 LimitNPROC=1048576
 LimitCORE=infinity
-MountFlags=slave
 TimeoutStartSec=1min
+Restart=on-abnormal
 
 [Install]
 WantedBy=multi-user.target

roles/kubernetes/master/handlers/main.yml

@@ -1,4 +1,35 @@
 ---
-- name: restart kube-apiserver
+- name: Master | restart kubelet
-  set_fact:
+  command: /bin/true
-    restart_apimaster: True
+  notify:
+    - Kubelet | reload systemd
+    - Kubelet | reload kubelet
+
+- name: wait for master static pods
+  command: /bin/true
+  notify:
+    - wait for kube-scheduler
+    - wait for kube-controller-manager
+
+- name: Master | reload systemd
+  command: systemctl daemon-reload
+  when: ansible_service_mgr == "systemd"
+
+- name: Master | reload kubelet
+  service:
+    name: kubelet
+    state: restarted
+
+- name: wait for kube-scheduler
+  uri: url=http://localhost:10251/healthz
+  register: scheduler_result
+  until: scheduler_result.status == 200
+  retries: 15
+  delay: 5
+
+- name: wait for kube-controller-manager
+  uri: url=http://localhost:10252/healthz
+  register: controller_manager_result
+  until: controller_manager_result.status == 200
+  retries: 15
+  delay: 5

roles/kubernetes/master/tasks/main.yml

@@ -19,12 +19,7 @@
     src: manifests/kube-apiserver.manifest.j2
     dest: "{{ kube_manifest_dir }}/kube-apiserver.manifest"
   register: apiserver_manifest
-
+  notify: Master | restart kubelet
-- name: restart kubelet
-  service:
-    name: kubelet
-    state: restarted
-  when: apiserver_manifest.changed
 
 - name: wait for the apiserver to be running
   uri: url=http://localhost:8080/healthz
@@ -58,8 +53,10 @@
   template:
     src: manifests/kube-controller-manager.manifest.j2
     dest: "{{ kube_manifest_dir }}/kube-controller-manager.manifest"
+  notify: wait for kube-controller-manager
 
 - name: Write kube-scheduler manifest
   template:
     src: manifests/kube-scheduler.manifest.j2
     dest: "{{ kube_manifest_dir }}/kube-scheduler.manifest"
+  notify: wait for kube-scheduler

roles/kubernetes/node/templates/kubelet.service.j2

@@ -2,9 +2,11 @@
 Description=Kubernetes Kubelet Server
 Documentation=https://github.com/GoogleCloudPlatform/kubernetes
 {% if kube_network_plugin is defined and kube_network_plugin == "calico" %}
-After=docker.service calico-node.service
+After=docker.service docker.socket calico-node.service
+Wants=docker.socket calico-node.service
 {% else %}
 After=docker.service
+Wants=docker.socket
 {% endif %}
 
 [Service]

roles/network_plugin/calico/tasks/main.yml

@@ -10,13 +10,6 @@
     - restart docker
   when: ansible_os_family != "CoreOS"
 
-- name: Calico | Write docker.service systemd file
-  template:
-    src: systemd-docker.service
-    dest: /lib/systemd/system/docker.service
-  notify: restart docker
-  when: ansible_service_mgr == "systemd" and ansible_os_family != "CoreOS"
-
 - meta: flush_handlers
 
 - name: Calico | Install calicoctl container script

roles/network_plugin/calico/templates/calico-node.service.j2

@@ -1,8 +1,8 @@
 [Unit]
 Description=Calico per-node agent
 Documentation=https://github.com/projectcalico/calico-docker
-After=docker.service etcd-proxy.service
+After=docker.service docker.socket etcd-proxy.service
-Wants=docker.socket
+Wants=docker.socket etcd-proxy.service
 
 [Service]
 User=root

roles/network_plugin/flannel/handlers/main.yml

@@ -4,22 +4,6 @@
   ignore_errors: yes
   notify: restart docker
 
-- name: restart docker
-  command: /bin/true
-  notify:
-    - Flannel | reload systemd
-    - Flannel | reload docker
-    - Flannel | reload kubelet
-
-- name : Flannel | reload systemd
-  shell: systemctl daemon-reload
-  when: ansible_service_mgr == "systemd"
-
-- name: Flannel | reload docker
-  service:
-    name: docker
-    state: restarted
-
 - name: Flannel | reload kubelet
   service:
     name: kubelet

roles/network_plugin/flannel/tasks/main.yml

@@ -50,11 +50,4 @@
     state: link
   when: ansible_os_family == "CoreOS"
 
-- name: Flannel | Write docker.service systemd file
-  template:
-    src: systemd-docker.service
-    dest: /lib/systemd/system/docker.service
-  notify: restart docker
-  when: ansible_service_mgr == "systemd" and ansible_os_family != "CoreOS"
-
 - meta: flush_handlers

roles/network_plugin/flannel/templates/systemd-docker.service

@@ -1,28 +0,0 @@
-[Unit]
-Description=Docker Application Container Engine
-Documentation=http://docs.docker.com
-{% if ansible_os_family == "RedHat" %}
-After=network.target
-Wants=docker-storage-setup.service
-{% elif ansible_os_family == "Debian" %}
-After=network.target docker.socket
-Requires=docker.socket
-{% endif %}
-
-[Service]
-Type=notify
-EnvironmentFile=-/etc/default/docker
-Environment=GOTRACEBACK=crash
-ExecStart=/usr/bin/docker daemon \
-          $OPTIONS \
-          $DOCKER_STORAGE_OPTIONS \
-          $DOCKER_NETWORK_OPTIONS \
-          $INSECURE_REGISTRY
-LimitNOFILE=1048576
-LimitNPROC=1048576
-LimitCORE=infinity
-MountFlags=slave
-TimeoutStartSec=1min
-
-[Install]
-WantedBy=multi-user.target

roles/network_plugin/meta/main.yml

@@ -6,3 +6,4 @@ dependencies:
    when: kube_network_plugin == 'flannel'
  - role: network_plugin/weave
    when: kube_network_plugin == 'weave'
+ - role: docker

roles/network_plugin/weave/handlers/main.yml

@@ -1,10 +1,4 @@
 ---
-- name: Weave | restart docker
-  command: /bin/true
-  notify:
-    - Weave | reload systemd
-    - Weave | reload docker
-
 - name: restart weave
   command: /bin/true
   notify:
@@ -27,11 +21,6 @@
     - Weave | reload systemd
     - reload weaveexpose
 
-- name: Weave | reload docker
-  service:
-    name: docker
-    state: restarted
-
 - name: reload weave
   service:
     name: weave

roles/network_plugin/weave/tasks/main.yml

@@ -7,14 +7,7 @@
     group: root
     mode: 0644
   notify:
-    - Weave | restart docker
+    - restart docker
-
-- name: Write docker.service systemd file
-  template:
-    src: systemd-docker.service
-    dest: /lib/systemd/system/docker.service
-  notify: Weave | restart docker
-  when: ansible_service_mgr == "systemd" and ansible_os_family != "CoreOS"
 
 - name: Weave | Install weave
   command: rsync -piu "{{ local_release_dir }}/weave/bin/weave" "{{ bin_dir }}/weave"

roles/network_plugin/weave/templates/systemd-docker.service

@@ -1,28 +0,0 @@
-[Unit]
-Description=Docker Application Container Engine
-Documentation=http://docs.docker.com
-{% if ansible_os_family == "RedHat" %}
-After=network.target
-Wants=docker-storage-setup.service
-{% elif ansible_os_family == "Debian" %}
-After=network.target docker.socket
-Requires=docker.socket
-{% endif %}
-
-[Service]
-Type=notify
-EnvironmentFile=-/etc/default/docker
-Environment=GOTRACEBACK=crash
-ExecStart=/usr/bin/docker daemon \
-          $OPTIONS \
-          $DOCKER_STORAGE_OPTIONS \
-          $DOCKER_NETWORK_OPTIONS \
-          $INSECURE_REGISTRY
-LimitNOFILE=1048576
-LimitNPROC=1048576
-LimitCORE=infinity
-MountFlags=slave
-TimeoutStartSec=1min
-
-[Install]
-WantedBy=multi-user.target

roles/network_plugin/weave/templates/weave.service.j2

@@ -1,8 +1,8 @@
 [Unit]
 Description=Weave Network
 Documentation=http://docs.weave.works/weave/latest_release/
-Requires=docker.service
+Wants=docker.socket
-After=docker.service
+After=docker.service docker.socket
 
 [Service]
 EnvironmentFile=-/etc/weave.env

roles/network_plugin/weave/templates/weaveexpose.service.j2

@@ -1,9 +1,7 @@
 [Unit]
 Documentation=http://docs.weave.works/
-Requires=docker.service
+Wants=docker.socket weave.service
-Requires=weave.service
+After=docker.service docker.socket weave.service
-After=weave.service
-After=docker.service
 
 [Service]
 Type=oneshot

roles/network_plugin/weave/templates/weaveproxy.service.j2

@@ -1,8 +1,8 @@
 [Unit]
 Description=Weave proxy for Docker API
 Documentation=http://docs.weave.works/
-Requires=docker.service
+Wants=docker.socket
-After=docker.service
+After=docker.service docker.socket
 
 [Service]
 EnvironmentFile=-/etc/weave.%H.env