Merge pull request #2827 from abutcher/BZ1377619

Allow ansible to continue when a node is unaccessible or fails.

3 files changed, 51 insertions, 41 deletions

diff --git a/playbooks/common/openshift-cluster/initialize_facts.yml b/playbooks/common/openshift-cluster/initialize_facts.yml
index 6d83d2527..18f99728c 100644
--- a/playbooks/common/openshift-cluster/initialize_facts.yml
+++ b/playbooks/common/openshift-cluster/initialize_facts.yml
@@ -1,7 +1,11 @@
 ---
+- name: Ensure that all non-node hosts are accessible
+  hosts: oo_masters_to_config:oo_etcd_to_config:oo_lb_to_config:oo_nfs_to_config
+  any_errors_fatal: true
+  tasks:
+
 - name: Initialize host facts
   hosts: oo_all_hosts
-  any_errors_fatal: true
   roles:
   - openshift_facts
   tasks:
diff --git a/playbooks/common/openshift-node/config.yml b/playbooks/common/openshift-node/config.yml
index 4824eeef3..e28da5713 100644
--- a/playbooks/common/openshift-node/config.yml
+++ b/playbooks/common/openshift-node/config.yml
@@ -139,6 +139,8 @@
   - role: nuage_node
     when: openshift.common.use_nuage | bool
   - role: nickhammond.logrotate
+  - role: openshift_manage_node
+    openshift_master_host: "{{ groups.oo_first_master.0 }}"
   tasks:
   - name: Create group for deployment type
     group_by: key=oo_nodes_deployment_type_{{ openshift.common.deployment_type }}
@@ -152,35 +154,3 @@
   tasks:
   - file: name={{ mktemp.stdout }} state=absent
     changed_when: False
-
-- name: Set node schedulability
-  hosts: oo_first_master
-  vars:
-    openshift_nodes: "{{ groups.oo_nodes_to_config | default([]) }}"
-  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 --silent --tlsv1.2
-      {% if openshift.common.version_gte_3_2_or_1_2 | bool %}
-      --cacert {{ openshift.common.config_base }}/master/ca-bundle.crt
-      {% else %}
-      --cacert {{ openshift.common.config_base }}/master/ca.crt
-      {% endif %}
-      {{ openshift.master.api_url }}/healthz/ready
-    args:
-      # Disables the following warning:
-      # Consider using get_url or uri module rather than running curl
-      warn: no
-    register: api_available_output
-    until: api_available_output.stdout == 'ok'
-    retries: 120
-    delay: 1
-    changed_when: false
-    when: openshift.common.is_containerized | bool
-  roles:
-  - openshift_manage_node
diff --git a/roles/openshift_manage_node/tasks/main.yml b/roles/openshift_manage_node/tasks/main.yml
index 28e4e46e9..88cdd2d89 100644
--- a/roles/openshift_manage_node/tasks/main.yml
+++ b/roles/openshift_manage_node/tasks/main.yml
@@ -3,18 +3,51 @@
   command: mktemp -d /tmp/openshift-ansible-XXXXXX
   register: mktemp
   changed_when: False
+  delegate_to: "{{ openshift_master_host }}"
+  run_once: true
 
 - set_fact:
     openshift_manage_node_kubeconfig: "{{ mktemp.stdout }}/admin.kubeconfig"
+  delegate_to: "{{ openshift_master_host }}"
+  run_once: true
 
 - name: Copy the admin client config(s)
   command: >
     cp {{ openshift.common.config_base }}/master/admin.kubeconfig {{ openshift_manage_node_kubeconfig }}
   changed_when: False
+  delegate_to: "{{ openshift_master_host }}"
+  run_once: true
+
+# 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 --silent --tlsv1.2
+    {% if openshift.common.version_gte_3_2_or_1_2 | bool %}
+    --cacert {{ openshift.common.config_base }}/master/ca-bundle.crt
+    {% else %}
+    --cacert {{ openshift.common.config_base }}/master/ca.crt
+    {% endif %}
+    {{ openshift_node_master_api_url }}/healthz/ready
+  args:
+    # Disables the following warning:
+    # Consider using get_url or uri module rather than running curl
+    warn: no
+  register: api_available_output
+  until: api_available_output.stdout == 'ok'
+  retries: 120
+  delay: 1
+  changed_when: false
+  when: openshift.common.is_containerized | bool
+  delegate_to: "{{ openshift_master_host }}"
+  run_once: true
 
 - name: Wait for Node Registration
   command: >
-    {{ openshift.common.client_binary }} get node {{ hostvars[item].openshift.node.nodename }}
+    {{ openshift.common.client_binary }} get node {{ openshift.node.nodename }}
     --config={{ openshift_manage_node_kubeconfig }}
     -n default
   register: omd_get_node
@@ -22,26 +55,29 @@
   retries: 50
   delay: 5
   changed_when: false
-  with_items: "{{ openshift_nodes }}"
+  when: "'nodename' in openshift.node"
+  delegate_to: "{{ openshift_master_host }}"
 
 - name: Set node schedulability
   command: >
-    {{ openshift.common.client_binary }} adm manage-node {{ hostvars[item].openshift.node.nodename }} --schedulable={{ 'true' if hostvars[item].openshift.node.schedulable | bool else 'false' }}
+    {{ openshift.common.client_binary }} adm manage-node {{ openshift.node.nodename }} --schedulable={{ 'true' if openshift.node.schedulable | bool else 'false' }}
     --config={{ openshift_manage_node_kubeconfig }}
     -n default
-  with_items: "{{ openshift_nodes }}"
-  when: hostvars[item].openshift.node.nodename is defined
+  when: "'nodename' in openshift.node"
+  delegate_to: "{{ openshift_master_host }}"
 
 - name: Label nodes
   command: >
-    {{ openshift.common.client_binary }} label --overwrite node {{ hostvars[item].openshift.node.nodename }} {{ hostvars[item].openshift.node.labels | oo_combine_dict  }}
+    {{ openshift.common.client_binary }} label --overwrite node {{ openshift.node.nodename }} {{ openshift.node.labels | oo_combine_dict  }}
     --config={{ openshift_manage_node_kubeconfig }}
     -n default
-  with_items: "{{ openshift_nodes }}"
-  when: hostvars[item].openshift.node.nodename is defined and 'labels' in hostvars[item].openshift.node and hostvars[item].openshift.node.labels != {}
+  when: "'nodename' in openshift.node and 'labels' in openshift.node and openshift.node.labels != {}"
+  delegate_to: "{{ openshift_master_host }}"
 
 - name: Delete temp directory
   file:
     name: "{{ mktemp.stdout }}"
     state: absent
   changed_when: False
+  delegate_to: "{{ openshift_master_host }}"
+  run_once: true