Merge pull request #608 from abutcher/native-ha

Native Support for Multi-Master HA

9 files changed, 195 insertions, 9 deletions

diff --git a/roles/openshift_master/handlers/main.yml b/roles/openshift_master/handlers/main.yml
index 37028e0f6..4b9500cbd 100644
--- a/roles/openshift_master/handlers/main.yml
+++ b/roles/openshift_master/handlers/main.yml
@@ -2,3 +2,13 @@
 - name: restart master
   service: name={{ openshift.common.service_type }}-master state=restarted
   when: (not openshift_master_ha | bool) and (not master_service_status_changed | default(false))
+
+- name: restart master api
+  service: name={{ openshift.common.service_type }}-master-api state=restarted
+  when: (openshift_master_ha | bool) and (not master_api_service_status_changed | default(false)) and openshift.master.cluster_method == 'native'
+
+# TODO: need to fix up ignore_errors here
+- name: restart master controllers
+  service: name={{ openshift.common.service_type }}-master-controllers state=restarted
+  when: (openshift_master_ha | bool) and (not master_controllers_service_status_changed | default(false)) and openshift.master.cluster_method == 'native'
+  ignore_errors: yes
diff --git a/roles/openshift_master/tasks/main.yml b/roles/openshift_master/tasks/main.yml
index 3a886935f..be77fce4a 100644
--- a/roles/openshift_master/tasks/main.yml
+++ b/roles/openshift_master/tasks/main.yml
@@ -9,16 +9,22 @@
   when: openshift_master_oauth_grant_method is defined
 
 - fail:
+    msg: "openshift_master_cluster_method must be set to either 'native' or 'pacemaker' for multi-master installations"
+  when: openshift_master_ha | bool and ((openshift_master_cluster_method is not defined) or (openshift_master_cluster_method is defined and openshift_master_cluster_method not in ["native", "pacemaker"]))
+- fail:
+    msg: "'native' high availability is not supported for the requested OpenShift version"
+  when: openshift_master_ha | bool and openshift_master_cluster_method == "native" and not openshift.common.version_greater_than_3_1_or_1_1 | bool
+- fail:
     msg: "openshift_master_cluster_password must be set for multi-master installations"
-  when: openshift_master_ha | bool and not openshift.master.cluster_defer_ha | bool and openshift_master_cluster_password is not defined
+  when: openshift_master_ha | bool and openshift_master_cluster_method == "pacemaker" and (openshift_master_cluster_password is not defined or not openshift_master_cluster_password)
 
 - name: Set master facts
   openshift_facts:
     role: master
     local_facts:
+      cluster_method: "{{ openshift_master_cluster_method | default(None) }}"
       cluster_hostname: "{{ openshift_master_cluster_hostname | default(None) }}"
       cluster_public_hostname: "{{ openshift_master_cluster_public_hostname | default(None) }}"
-      cluster_defer_ha: "{{ openshift_master_cluster_defer_ha | default(None) }}"
       debug_level: "{{ openshift_master_debug_level | default(openshift.common.debug_level) }}"
       api_port: "{{ openshift_master_api_port | default(None) }}"
       api_url: "{{ openshift_master_api_url | default(None) }}"
@@ -41,6 +47,8 @@
       portal_net: "{{ openshift_master_portal_net | default(None) }}"
       session_max_seconds: "{{ openshift_master_session_max_seconds | default(None) }}"
       session_name: "{{ openshift_master_session_name | default(None) }}"
+      session_auth_secrets: "{{ openshift_master_session_auth_secrets | default(None) }}"
+      session_encryption_secrets: "{{ openshift_master_session_encryption_secrets | default(None) }}"
       session_secrets_file: "{{ openshift_master_session_secrets_file | default(None) }}"
       access_token_max_seconds: "{{ openshift_master_access_token_max_seconds | default(None) }}"
       auth_token_max_seconds: "{{ openshift_master_auth_token_max_seconds | default(None) }}"
@@ -63,6 +71,8 @@
       controller_args: "{{ osm_controller_args | default(None) }}"
       infra_nodes: "{{ num_infra | default(None) }}"
       disabled_features: "{{ osm_disabled_features | default(None) }}"
+      master_count: "{{ openshift_master_count | default(None) }}"
+      controller_lease_ttl: "{{ osm_controller_lease_ttl | default(None) }}"
 
 - name: Install Master package
   yum: pkg={{ openshift.common.service_type }}-master{{ openshift_version  }} state=present
@@ -77,7 +87,7 @@
       domain: cluster.local
   when: openshift.master.embedded_dns
 
-- name: Create config parent directory if it doesn't exist
+- name: Create config parent directory if it does not exist
   file:
     path: "{{ openshift_master_config_dir }}"
     state: directory
@@ -90,6 +100,8 @@
     creates: "{{ openshift_master_policy }}"
   notify:
   - restart master
+  - restart master api
+  - restart master controllers
 
 - name: Create the scheduler config
   template:
@@ -98,6 +110,8 @@
     backup: true
   notify:
   - restart master
+  - restart master api
+  - restart master controllers
 
 - name: Install httpd-tools if needed
   yum: pkg=httpd-tools state=present
@@ -120,6 +134,39 @@
   when: item.kind == 'HTPasswdPasswordIdentityProvider'
   with_items: openshift.master.identity_providers
 
+# workaround for missing systemd unit files for controllers/api
+- name: Create the api service file
+  template:
+    src: atomic-openshift-master-api.service.j2
+    dest: /usr/lib/systemd/system/{{ openshift.common.service_type }}-master-api.service
+    force: no
+- name: Create the controllers service file
+  template:
+    src: atomic-openshift-master-controllers.service.j2
+    dest: /usr/lib/systemd/system/{{ openshift.common.service_type }}-master-controllers.service
+    force: no
+- name: Create the api env file
+  template:
+    src: atomic-openshift-master-api.j2
+    dest: /etc/sysconfig/{{ openshift.common.service_type }}-master-api
+    force: no
+- name: Create the controllers env file
+  template:
+    src: atomic-openshift-master-controllers.j2
+    dest: /etc/sysconfig/{{ openshift.common.service_type }}-master-controllers
+    force: no
+- command: systemctl daemon-reload
+# end workaround for missing systemd unit files
+
+- name: Create session secrets file
+  template:
+    dest: "{{ openshift.master.session_secrets_file }}"
+    src: sessionSecretsFile.yaml.v1.j2
+    force: no
+  notify:
+  - restart master
+  - restart master api
+
 # TODO: add the validate parameter when there is a validation command to run
 - name: Create master config
   template:
@@ -128,12 +175,15 @@
     backup: true
   notify:
   - restart master
+  - restart master api
+  - restart master controllers
 
 - name: Configure master settings
   lineinfile:
     dest: /etc/sysconfig/{{ openshift.common.service_type }}-master
     regexp: "{{ item.regex }}"
     line: "{{ item.line }}"
+    create: yes
   with_items:
     - regex: '^OPTIONS='
       line: "OPTIONS=--loglevel={{ openshift.master.debug_level }}"
@@ -142,6 +192,32 @@
   notify:
   - restart master
 
+- name: Configure master api settings
+  lineinfile:
+    dest: /etc/sysconfig/{{ openshift.common.service_type }}-master-api
+    regexp: "{{ item.regex }}"
+    line: "{{ item.line }}"
+  with_items:
+    - regex: '^OPTIONS='
+      line: "OPTIONS=--loglevel={{ openshift.master.debug_level }} --listen=https://0.0.0.0:8443 --master=https://{{ openshift.common.ip }}:8443"
+    - regex: '^CONFIG_FILE='
+      line: "CONFIG_FILE={{ openshift_master_config_file }}"
+  notify:
+  - restart master api
+
+- name: Configure master controller settings
+  lineinfile:
+    dest: /etc/sysconfig/{{ openshift.common.service_type }}-master-controllers
+    regexp: "{{ item.regex }}"
+    line: "{{ item.line }}"
+  with_items:
+    - regex: '^OPTIONS='
+      line: "OPTIONS=--loglevel={{ openshift.master.debug_level }} --listen=https://0.0.0.0:8444"
+    - regex: '^CONFIG_FILE='
+      line: "CONFIG_FILE={{ openshift_master_config_file }}"
+  notify:
+  - restart master controllers
+
 - name: Start and enable master
   service: name={{ openshift.common.service_type }}-master enabled=yes state=started
   when: not openshift_master_ha | bool
@@ -149,15 +225,37 @@
 
 - set_fact:
     master_service_status_changed = start_result | changed
+  when: not openshift_master_ha | bool
+
+- name: Start and enable master api
+  service: name={{ openshift.common.service_type }}-master-api enabled=yes state=started
+  when: openshift_master_ha | bool and openshift.master.cluster_method == 'native'
+  register: start_result
+
+- set_fact:
+    master_api_service_status_changed = start_result | changed
+  when: openshift_master_ha | bool and openshift.master.cluster_method == 'native'
+
+# TODO: fix the ugly workaround of setting ignore_errors
+#       the controllers service tries to start even if it is already started
+- name: Start and enable master controller
+  service: name={{ openshift.common.service_type }}-master-controllers enabled=yes state=started
+  when: openshift_master_ha | bool and openshift.master.cluster_method == 'native'
+  register: start_result
+  ignore_errors: yes
+
+- set_fact:
+    master_controllers_service_status_changed = start_result | changed
+  when: openshift_master_ha | bool and openshift.master.cluster_method == 'native'
 
 - name: Install cluster packages
   yum: pkg=pcs state=present
-  when: openshift_master_ha | bool and not openshift.master.cluster_defer_ha | bool
+  when: openshift_master_ha | bool and openshift.master.cluster_method == 'pacemaker'
   register: install_result
 
 - name: Start and enable cluster service
   service: name=pcsd enabled=yes state=started
-  when: openshift_master_ha | bool and not openshift.master.cluster_defer_ha | bool
+  when: openshift_master_ha | bool and openshift.master.cluster_method == 'pacemaker'
 
 - name: Set the cluster user password
   shell: echo {{ openshift_master_cluster_password | quote }} | passwd --stdin hacluster
diff --git a/roles/openshift_master/templates/atomic-openshift-master-api.j2 b/roles/openshift_master/templates/atomic-openshift-master-api.j2
new file mode 100644
index 000000000..205934248
--- /dev/null
+++ b/roles/openshift_master/templates/atomic-openshift-master-api.j2
@@ -0,0 +1,9 @@
+OPTIONS=
+CONFIG_FILE={{ openshift_master_config_dir }}/master-config.yaml
+
+# Proxy configuration
+# Origin uses standard HTTP_PROXY environment variables. Be sure to set
+# NO_PROXY for your master
+#NO_PROXY=master.example.com
+#HTTP_PROXY=http://USER:PASSWORD@IPADDR:PORT
+#HTTPS_PROXY=https://USER:PASSWORD@IPADDR:PORT
diff --git a/roles/openshift_master/templates/atomic-openshift-master-api.service.j2 b/roles/openshift_master/templates/atomic-openshift-master-api.service.j2
new file mode 100644
index 000000000..ba19fb348
--- /dev/null
+++ b/roles/openshift_master/templates/atomic-openshift-master-api.service.j2
@@ -0,0 +1,21 @@
+[Unit]
+Description=Atomic OpenShift Master API
+Documentation=https://github.com/openshift/origin
+After=network.target
+After=etcd.service
+Before={{ openshift.common.service_type }}-node.service
+Requires=network.target
+
+[Service]
+Type=notify
+EnvironmentFile=/etc/sysconfig/{{ openshift.common.service_type }}-master-api
+Environment=GOTRACEBACK=crash
+ExecStart=/usr/bin/openshift start master api --config=${CONFIG_FILE} $OPTIONS
+LimitNOFILE=131072
+LimitCORE=infinity
+WorkingDirectory={{ openshift.common.data_dir }}
+SyslogIdentifier=atomic-openshift-master-api
+
+[Install]
+WantedBy=multi-user.target
+WantedBy={{ openshift.common.service_type }}-node.service
diff --git a/roles/openshift_master/templates/atomic-openshift-master-controllers.j2 b/roles/openshift_master/templates/atomic-openshift-master-controllers.j2
new file mode 100644
index 000000000..205934248
--- /dev/null
+++ b/roles/openshift_master/templates/atomic-openshift-master-controllers.j2
@@ -0,0 +1,9 @@
+OPTIONS=
+CONFIG_FILE={{ openshift_master_config_dir }}/master-config.yaml
+
+# Proxy configuration
+# Origin uses standard HTTP_PROXY environment variables. Be sure to set
+# NO_PROXY for your master
+#NO_PROXY=master.example.com
+#HTTP_PROXY=http://USER:PASSWORD@IPADDR:PORT
+#HTTPS_PROXY=https://USER:PASSWORD@IPADDR:PORT
diff --git a/roles/openshift_master/templates/atomic-openshift-master-controllers.service.j2 b/roles/openshift_master/templates/atomic-openshift-master-controllers.service.j2
new file mode 100644
index 000000000..8952c86ef
--- /dev/null
+++ b/roles/openshift_master/templates/atomic-openshift-master-controllers.service.j2
@@ -0,0 +1,22 @@
+[Unit]
+Description=Atomic OpenShift Master Controllers
+Documentation=https://github.com/openshift/origin
+After=network.target
+After={{ openshift.common.service_type }}-master-api.service
+Before={{ openshift.common.service_type }}-node.service
+Requires=network.target
+
+[Service]
+Type=notify
+EnvironmentFile=/etc/sysconfig/{{ openshift.common.service_type }}-master-controllers
+Environment=GOTRACEBACK=crash
+ExecStart=/usr/bin/openshift start master controllers --config=${CONFIG_FILE} $OPTIONS
+LimitNOFILE=131072
+LimitCORE=infinity
+WorkingDirectory={{ openshift.common.data_dir }}
+SyslogIdentifier={{ openshift.common.service_type }}-master-controllers
+Restart=on-failure
+
+[Install]
+WantedBy=multi-user.target
+WantedBy={{ openshift.common.service_type }}-node.service
diff --git a/roles/openshift_master/templates/master.yaml.v1.j2 b/roles/openshift_master/templates/master.yaml.v1.j2
index 9547a6945..d4a6590ea 100644
--- a/roles/openshift_master/templates/master.yaml.v1.j2
+++ b/roles/openshift_master/templates/master.yaml.v1.j2
@@ -10,13 +10,18 @@ assetConfig:
   publicURL: {{ openshift.master.public_console_url }}/
   servingInfo:
     bindAddress: {{ openshift.master.bind_addr }}:{{ openshift.master.console_port }}
+    bindNetwork: tcp4
     certFile: master.server.crt
     clientCA: ""
     keyFile: master.server.key
     maxRequestsInFlight: 0
     requestTimeoutSeconds: 0
+{% if openshift_master_ha | bool %}
+controllerLeaseTTL: {{ openshift.master.controller_lease_ttl | default('30') }}
+{% endif %}
+controllers: '*'
 corsAllowedOrigins:
-{% for origin in ['127.0.0.1', 'localhost', openshift.common.hostname, openshift.common.ip, openshift.common.public_hostname, openshift.common.public_ip] | unique %}
+{% for origin in ['127.0.0.1', 'localhost', openshift.common.ip, openshift.common.public_ip] | union(openshift.common.all_hostnames) | unique %}
   - {{ origin }}
 {% endfor %}
 {% for custom_origin in openshift.master.custom_cors_origins | default("") %}
@@ -29,8 +34,10 @@ corsAllowedOrigins:
 disabledFeatures: {{ openshift.master.disabled_features | to_json }}
 {% endif %}
 {% if openshift.master.embedded_dns | bool %}
+disabledFeatures: null
 dnsConfig:
   bindAddress: {{ openshift.master.bind_addr }}:{{ openshift.master.dns_port }}
+  bindNetwork: tcp4
 {% endif %}
 etcdClientInfo:
   ca: {{ "ca.crt" if (openshift.master.embedded_etcd | bool) else "master.etcd-ca.crt" }}
@@ -80,9 +87,8 @@ kubernetesMasterConfig:
   - v1
   apiServerArguments: {{ api_server_args if api_server_args is defined else 'null' }}
   controllerArguments: {{ controller_args if controller_args is defined else 'null' }}
-{# TODO: support overriding masterCount #}
-  masterCount: 1
-  masterIP: ""
+  masterCount: {{ openshift.master.master_count }}
+  masterIP: {{ openshift.common.ip }}
   podEvictionTimeout: ""
   proxyClientInfo:
     certFile: master.proxy-client.crt
@@ -106,6 +112,7 @@ networkConfig:
 # serviceNetworkCIDR must match kubernetesMasterConfig.servicesSubnet
   serviceNetworkCIDR: {{ openshift.master.portal_net }}
 {% include 'v1_partials/oauthConfig.j2' %}
+pauseControllers: false
 policyConfig:
   bootstrapPolicyFile: {{ openshift_master_policy }}
   openshiftInfrastructureNamespace: openshift-infra
@@ -121,6 +128,7 @@ projectConfig:
 routingConfig:
   subdomain:  "{{ openshift.master.default_subdomain | default("") }}"
 serviceAccountConfig:
+  limitSecretReferences: false
   managedNames:
   - default
   - builder
@@ -131,6 +139,7 @@ serviceAccountConfig:
   - serviceaccounts.public.key
 servingInfo:
   bindAddress: {{ openshift.master.bind_addr }}:{{ openshift.master.api_port }}
+  bindNetwork: tcp4
   certFile: master.server.crt
   clientCA: ca.crt
   keyFile: master.server.key
diff --git a/roles/openshift_master/templates/sessionSecretsFile.yaml.v1.j2 b/roles/openshift_master/templates/sessionSecretsFile.yaml.v1.j2
new file mode 100644
index 000000000..d12d9db90
--- /dev/null
+++ b/roles/openshift_master/templates/sessionSecretsFile.yaml.v1.j2
@@ -0,0 +1,7 @@
+apiVersion: v1
+kind: SessionSecrets
+secrets:
+{% for secret in openshift_master_session_auth_secrets %}
+- authentication: "{{ openshift_master_session_auth_secrets[loop.index0] }}"
+  encryption: "{{ openshift_master_session_encryption_secrets[loop.index0] }}"
+{% endfor %}
diff --git a/roles/openshift_master/vars/main.yml b/roles/openshift_master/vars/main.yml
index ecdb4f883..534465451 100644
--- a/roles/openshift_master/vars/main.yml
+++ b/roles/openshift_master/vars/main.yml
@@ -2,6 +2,7 @@
 openshift_master_config_dir: "{{ openshift.common.config_base }}/master"
 openshift_master_config_file: "{{ openshift_master_config_dir }}/master-config.yaml"
 openshift_master_scheduler_conf: "{{ openshift_master_config_dir }}/scheduler.json"
+openshift_master_session_secrets_file: "{{ openshift_master_config_dir }}/session-secrets.yaml"
 openshift_master_policy: "{{ openshift_master_config_dir }}/policy.json"
 openshift_version: "{{ openshift_pkg_version | default('') }}"