Merge pull request #1159 from abutcher/wait-for-api

API Verification

4 files changed, 47 insertions, 1 deletions

diff --git a/playbooks/common/openshift-master/config.yml b/playbooks/common/openshift-master/config.yml
index 4ecdf2a0c..0df03f194 100644
--- a/playbooks/common/openshift-master/config.yml
+++ b/playbooks/common/openshift-master/config.yml
@@ -313,6 +313,7 @@
 
 - name: Configure master instances
   hosts: oo_masters_to_config
+  any_errors_fatal: true
   serial: 1
   vars:
     sync_tmpdir: "{{ hostvars.localhost.g_master_mktemp.stdout }}"
diff --git a/playbooks/common/openshift-node/config.yml b/playbooks/common/openshift-node/config.yml
index 483a7768c..336cbed5e 100644
--- a/playbooks/common/openshift-node/config.yml
+++ b/playbooks/common/openshift-node/config.yml
@@ -215,6 +215,23 @@
                          | oo_collect('openshift.common.hostname') }}"
     openshift_node_vars: "{{ hostvars | oo_select_keys(groups['oo_nodes_to_config']) }}"
   pre_tasks:
-
+  # Necessary because when you're on a node that's also a master the master will be
+  # restarted after the node restarts docker and it will take up to 60 seconds for
+  # systemd to start the master again
+  - name: Wait for master API to become available before proceeding
+    # Using curl here since the uri module requires python-httplib2 and
+    # wait_for port doesn't provide health information.
+    command: >
+      curl -k --head --silent {{ openshift.master.api_url }}
+    register: api_available_output
+    until: api_available_output.stdout.find("200 OK") != -1
+    retries: 120
+    delay: 1
+    changed_when: false
+    when: openshift.common.is_containerized | bool
+  - fail:
+      msg: >
+        Unable to contact master API at {{ openshift.master.api_url }}
+    when: openshift.common.is_containerized | bool and api_available_output.stdout.find("200 OK") == -1
   roles:
   - openshift_manage_node
diff --git a/roles/openshift_master/handlers/main.yml b/roles/openshift_master/handlers/main.yml
index e1b95eda4..6b9992eea 100644
--- a/roles/openshift_master/handlers/main.yml
+++ b/roles/openshift_master/handlers/main.yml
@@ -2,11 +2,24 @@
 - 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) | bool))
+  notify: Verify API Server
 
 - 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) | bool)) and openshift.master.cluster_method == 'native'
+  notify: Verify API Server
 
 - 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) | bool)) and openshift.master.cluster_method == 'native'
+
+- name: Verify API Server
+  # Using curl here since the uri module requires python-httplib2 and
+  # wait_for port doesn't provide health information.
+  command: >
+    curl -k --head --silent {{ openshift.master.api_url }}
+  register: api_available_output
+  until: api_available_output.stdout.find("200 OK") != -1
+  retries: 120
+  delay: 1
+  changed_when: false
diff --git a/roles/openshift_master/tasks/main.yml b/roles/openshift_master/tasks/main.yml
index 3b46a0df4..150b76fc8 100644
--- a/roles/openshift_master/tasks/main.yml
+++ b/roles/openshift_master/tasks/main.yml
@@ -269,6 +269,7 @@
   service: name={{ openshift.common.service_type }}-master enabled=yes state=started
   when: not openshift_master_ha | bool
   register: start_result
+  notify: Verify API Server
 
 - name: Stop and disable non HA master when running HA
   service: name={{ openshift.common.service_type }}-master enabled=no state=stopped
@@ -287,6 +288,20 @@
     master_api_service_status_changed: "{{ start_result | changed }}"
   when: openshift_master_ha | bool and openshift.master.cluster_method == 'native'
 
+# A separate wait is required here for native HA since notifies will
+# be resolved after all tasks in the role.
+- name: Wait for API to become available
+  # Using curl here since the uri module requires python-httplib2 and
+  # wait_for port doesn't provide health information.
+  command: >
+    curl -k --head --silent {{ openshift.master.api_url }}
+  register: api_available_output
+  until: api_available_output.stdout.find("200 OK") != -1
+  retries: 120
+  delay: 1
+  changed_when: false
+  when: openshift_master_ha | bool and openshift.master.cluster_method == 'native' and master_api_service_status_changed | bool
+
 - 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'