Merge pull request #1945 from dgoodwin/upgrade33

openshift_release / version / upgrade improvements

1 files changed, 44 insertions, 69 deletions

diff --git a/playbooks/common/openshift-cluster/upgrades/v3_1_to_v3_2/upgrade.yml b/playbooks/common/openshift-cluster/upgrades/v3_1_to_v3_2/upgrade.yml
index 964257af5..0ea315d0a 100644
--- a/playbooks/common/openshift-cluster/upgrades/v3_1_to_v3_2/upgrade.yml
+++ b/playbooks/common/openshift-cluster/upgrades/v3_1_to_v3_2/upgrade.yml
@@ -3,49 +3,6 @@
 # The restart playbook should be run after this playbook completes.
 ###############################################################################
 
-- name: Upgrade docker
-  hosts: oo_masters_to_config:oo_nodes_to_config
-  roles:
-  - openshift_facts
-  tasks:
-  - include: docker_upgrade.yml
-    when: not openshift.common.is_atomic | bool
-  - name: Set post docker install facts
-    openshift_facts:
-      role: "{{ item.role }}"
-      local_facts: "{{ item.local_facts }}"
-    with_items:
-    - role: docker
-      local_facts:
-        openshift_image_tag: "v{{ g_new_version }}"
-        openshift_version: "{{ g_new_version }}"
-
-- name: Upgrade docker
-  hosts: oo_etcd_to_config
-  roles:
-  - openshift_facts
-  tasks:
-  # Upgrade docker when host is not atomic and host is not a non-containerized etcd node
-  - include: docker_upgrade.yml
-    when: not openshift.common.is_atomic | bool and not ('oo_etcd_to_config' in group_names and not openshift.common.is_containerized)
-
-# The cli image is used by openshift_docker_facts to determine the currently installed
-# version.  We need to explicitly pull the latest image to handle cases where
-# the locally cached 'latest' tag is older the g_new_version.
-- name: Download cli image
-  hosts: oo_masters_to_config:oo_nodes_to_config
-  roles:
-  - { role: openshift_docker_facts }
-  vars:
-    openshift_docker_hosted_registry_network: "{{ hostvars[groups.oo_first_master.0].openshift.common.portal_net }}"
-  tasks:
-  - name: Pull Images
-    command: >
-      docker pull {{ item }}:latest
-    with_items:
-    - "{{ openshift.common.cli_image }}"
-    when: openshift.common.is_containerized | bool
-
 ###############################################################################
 # Upgrade Masters
 ###############################################################################
@@ -62,7 +19,7 @@
   - include_vars: ../../../../../roles/openshift_master/vars/main.yml
 
   - name: Update systemd units
-    include: ../../../../../roles/openshift_master/tasks/systemd_units.yml openshift_version=v{{ g_new_version }}
+    include: ../../../../../roles/openshift_master/tasks/systemd_units.yml
 
 #  - name: Upgrade master configuration
 #    openshift_upgrade_config:
@@ -98,36 +55,55 @@
 ###############################################################################
 # Upgrade Nodes
 ###############################################################################
-- name: Upgrade nodes
-  hosts: oo_nodes_to_config
+
+# Here we handle all tasks that might require a node evac. (upgrading docker, and the node service)
+- name: Perform upgrades that may require node evacuation
+  hosts: oo_masters_to_config:oo_etcd_to_config:oo_nodes_to_config
   serial: 1
+  any_errors_fatal: true
   roles:
   - openshift_facts
   handlers:
   - include: ../../../../../roles/openshift_node/handlers/main.yml
   tasks:
-  - include: node_upgrade.yml
+  # TODO: To better handle re-trying failed upgrades, it would be nice to check if the node
+  # or docker actually needs an upgrade before proceeding. Perhaps best to save this until
+  # we merge upgrade functionality into the base roles and a normal config.yml playbook run.
+  - name: Mark unschedulable if host is a node
+    command: >
+      {{ openshift.common.admin_binary }} manage-node {{ openshift.common.hostname | lower }} --schedulable=false
+    delegate_to: "{{ groups.oo_first_master.0 }}"
+    when: inventory_hostname in groups.oo_nodes_to_config
 
-  - set_fact:
-      node_update_complete: True
+  - name: Evacuate Node for Kubelet upgrade
+    command: >
+      {{ openshift.common.admin_binary }} manage-node {{ openshift.common.hostname | lower }} --evacuate --force
+    delegate_to: "{{ groups.oo_first_master.0 }}"
+    when: inventory_hostname in groups.oo_nodes_to_config
+
+  # Only check if docker upgrade is required if docker_upgrade is not
+  # already set to False.
+  - include: ../docker/upgrade_check.yml
+    when: docker_upgrade is not defined or docker_upgrade | bool
+
+  - include: ../docker/upgrade.yml
+    when: l_docker_upgrade is defined and l_docker_upgrade | bool
+
+  - include: rpm_upgrade.yml
+    vars:
+       component: "node"
+       openshift_version: "{{ openshift_pkg_version | default('') }}"
+    when: inventory_hostname in groups.oo_nodes_to_config and not openshift.common.is_containerized | bool
+
+  - include: containerized_node_upgrade.yml
+    when: inventory_hostname in groups.oo_nodes_to_config and openshift.common.is_containerized | bool
+
+  - name: Set node schedulability
+    command: >
+      {{ openshift.common.admin_binary }} manage-node {{ openshift.common.hostname | lower }} --schedulable=true
+    delegate_to: "{{ groups.oo_first_master.0 }}"
+    when: inventory_hostname in groups.oo_nodes_to_config and openshift.node.schedulable | bool
 
-##############################################################################
-# Gate on nodes update
-##############################################################################
-- name: Gate on nodes update
-  hosts: localhost
-  connection: local
-  become: no
-  tasks:
-  - set_fact:
-      node_update_completed: "{{ hostvars
-                                 | oo_select_keys(groups.oo_nodes_to_config)
-                                 | oo_collect('inventory_hostname', {'node_update_complete': true}) }}"
-  - set_fact:
-      node_update_failed: "{{ groups.oo_nodes_to_config | difference(node_update_completed) }}"
-  - fail:
-      msg: "Upgrade cannot continue. The following nodes did not finish updating: {{ node_update_failed | join(',') }}"
-    when: node_update_failed | length > 0
 
 ###############################################################################
 # Reconcile Cluster Roles, Cluster Role Bindings and Security Context Constraints
@@ -136,12 +112,11 @@
 - name: Reconcile Cluster Roles and Cluster Role Bindings and Security Context Constraints
   hosts: oo_masters_to_config
   roles:
-  - { role: openshift_cli, openshift_image_tag: "v{{ g_new_version }}" }
+  - { role: openshift_cli }
   vars:
-    origin_reconcile_bindings: "{{ deployment_type == 'origin' and g_new_version | version_compare('1.0.6', '>') }}"
+    origin_reconcile_bindings: "{{ deployment_type == 'origin' and openshift_version | version_compare('1.0.6', '>') }}"
     ent_reconcile_bindings: true
     openshift_docker_hosted_registry_network: "{{ hostvars[groups.oo_first_master.0].openshift.common.portal_net }}"
-    upgrading: True
   tasks:
   - name: Verifying the correct commandline tools are available
     shell: grep {{ verify_upgrade_version }} {{ openshift.common.admin_binary}}