okd/playbooks/common/openshift-cluster/upgrades/v3_0_to_v3_1/upgrade.yml

---
###############################################################################
# Evaluate host groups and gather facts
###############################################################################
- name: Evaluate host groups
  include: ../../evaluate_groups.yml

- name: Load openshift_facts
  hosts: oo_masters_to_config:oo_nodes_to_config:oo_etcd_to_config:oo_lb_to_config
  roles:
  - openshift_facts

- name: Evaluate additional groups for upgrade
  hosts: localhost
  tasks:
  - name: Evaluate etcd_hosts_to_backup
    add_host:
      name: "{{ item }}"
      groups: etcd_hosts_to_backup
    with_items: groups.oo_etcd_to_config if groups.oo_etcd_to_config is defined and groups.oo_etcd_to_config | length > 0 else groups.oo_first_master


###############################################################################
# Pre-upgrade checks
###############################################################################
- name: Verify upgrade can proceed
  hosts: oo_first_master
  vars:
    openshift_master_ha: "{{ groups.oo_masters_to_config | length > 1 }}"
  gather_facts: no
  tasks:
  # Pacemaker is currently the only supported upgrade path for multiple masters
  - fail:
      msg: "openshift_master_cluster_method must be set to 'pacemaker'"
    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 != "pacemaker"))

  - fail:
      msg: >
        This upgrade is only supported for origin, openshift-enterprise, and online
        deployment types
    when: deployment_type not in ['origin','openshift-enterprise', 'online']

  - fail:
      msg: >
        openshift_pkg_version is {{ openshift_pkg_version }} which is not a
        valid version for a 3.1 upgrade
    when: openshift_pkg_version is defined and openshift_pkg_version.split('-',1).1 | version_compare('3.0.2.900','<')

  # If this script errors out ansible will show the default stdout/stderr
  # which contains details for the user:
  - script: ../files/pre-upgrade-check


- name: Verify upgrade can proceed
  hosts: oo_masters_to_config:oo_nodes_to_config
  tasks:
  - name: Clean package cache
    command: "{{ ansible_pkg_mgr }} clean all"

  - set_fact:
      g_new_service_name: "{{ 'origin' if deployment_type =='origin' else 'atomic-openshift' }}"

  - name: Determine available versions
    script: ../files/versions.sh {{ g_new_service_name }} openshift
    register: g_versions_result

  - set_fact:
      g_aos_versions: "{{ g_versions_result.stdout | from_yaml }}"

  - set_fact:
      g_new_version: "{{ g_aos_versions.curr_version.split('-', 1).0 if g_aos_versions.avail_version is none else g_aos_versions.avail_version.split('-', 1).0 }}"

  - fail:
      msg: This playbook requires Origin 1.0.6 or later
    when: deployment_type == 'origin' and g_aos_versions.curr_version | version_compare('1.0.6','<')

  - fail:
      msg: Atomic OpenShift 3.1 packages not found
    when: g_aos_versions.curr_version | version_compare('3.0.2.900','<') and (g_aos_versions.avail_version is none or g_aos_versions.avail_version | version_compare('3.0.2.900','<'))

  - set_fact:
      pre_upgrade_complete: True


##############################################################################
# Gate on pre-upgrade checks
##############################################################################
- name: Gate on pre-upgrade checks
  hosts: localhost
  vars:
    pre_upgrade_hosts: "{{ groups.oo_masters_to_config | union(groups.oo_nodes_to_config) }}"
  tasks:
  - set_fact:
      pre_upgrade_completed: "{{ hostvars
                                 | oo_select_keys(pre_upgrade_hosts)
                                 | oo_collect('inventory_hostname', {'pre_upgrade_complete': true}) }}"
  - set_fact:
      pre_upgrade_failed: "{{ pre_upgrade_hosts | difference(pre_upgrade_completed) }}"
  - fail:
      msg: "Upgrade cannot continue. The following hosts did not complete pre-upgrade checks: {{ pre_upgrade_failed | join(',') }}"
    when: pre_upgrade_failed | length > 0



###############################################################################
# Backup etcd
###############################################################################
- name: Backup etcd
  hosts: etcd_hosts_to_backup
  vars:
    embedded_etcd: "{{ openshift.master.embedded_etcd }}"
    timestamp: "{{ lookup('pipe', 'date +%Y%m%d%H%M%S') }}"
  roles:
  - openshift_facts
  tasks:
  # Ensure we persist the etcd role for this host in openshift_facts
  - openshift_facts:
      role: etcd
      local_facts: {}
    when: "'etcd' not in openshift"

  - stat: path=/var/lib/openshift
    register: var_lib_openshift

  - stat: path=/var/lib/origin
    register: var_lib_origin

  - name: Create origin symlink if necessary
    file: src=/var/lib/openshift/ dest=/var/lib/origin state=link
    when: var_lib_openshift.stat.exists == True and var_lib_origin.stat.exists == False

  # TODO: replace shell module with command and update later checks
  # We assume to be using the data dir for all backups.
  - name: Check available disk space for etcd backup
    shell: df --output=avail -k {{ openshift.common.data_dir }} | tail -n 1
    register: avail_disk

  # TODO: replace shell module with command and update later checks
  - name: Check current embedded etcd disk usage
    shell: du -k {{ openshift.etcd.etcd_data_dir }} | tail -n 1 | cut -f1
    register: etcd_disk_usage
    when: embedded_etcd | bool

  - name: Abort if insufficient disk space for etcd backup
    fail:
      msg: >
        {{ etcd_disk_usage.stdout }} Kb disk space required for etcd backup,
        {{ avail_disk.stdout }} Kb available.
    when: (embedded_etcd | bool) and (etcd_disk_usage.stdout|int > avail_disk.stdout|int)

  - name: Install etcd (for etcdctl)
    action: "{{ ansible_pkg_mgr }} name=etcd state=latest"

  - name: Generate etcd backup
    command: >
      etcdctl backup --data-dir={{ openshift.etcd.etcd_data_dir }}
      --backup-dir={{ openshift.common.data_dir }}/etcd-backup-{{ timestamp }}

  - set_fact:
      etcd_backup_complete: True

  - name: Display location of etcd backup
    debug:
      msg: "Etcd backup created in {{ openshift.common.data_dir }}/etcd-backup-{{ timestamp }}"


##############################################################################
# Gate on etcd backup
##############################################################################
- name: Gate on etcd backup
  hosts: localhost
  tasks:
  - set_fact:
      etcd_backup_completed: "{{ hostvars
                                 | oo_select_keys(groups.etcd_hosts_to_backup)
                                 | oo_collect('inventory_hostname', {'etcd_backup_complete': true}) }}"
  - set_fact:
      etcd_backup_failed: "{{ groups.etcd_hosts_to_backup | difference(etcd_backup_completed) }}"
  - fail:
      msg: "Upgrade cannot continue. The following hosts did not complete etcd backup: {{ etcd_backup_failed | join(',') }}"
    when: etcd_backup_failed | length > 0



###############################################################################
# Upgrade Masters
###############################################################################
- name: Create temp directory for syncing certs
  hosts: localhost
  gather_facts: no
  tasks:
  - name: Create local temp directory for syncing certs
    local_action: command mktemp -d /tmp/openshift-ansible-XXXXXXX
    register: g_master_mktemp
    changed_when: False

- name: Update deployment type
  hosts: oo_masters_to_config:oo_nodes_to_config:oo_etcd_to_config
  roles:
  - openshift_facts
  post_tasks:
  - openshift_facts:
      role: common
      local_facts:
        deployment_type: "{{ deployment_type }}"

- name: Update master facts
  hosts: oo_masters_to_config
  roles:
  - openshift_facts
  post_tasks:
  - openshift_facts:
      role: master
      local_facts:
        cluster_method: "{{ openshift_master_cluster_method | default(None) }}"

- name: Upgrade master packages and configuration
  hosts: oo_masters_to_config
  vars:
    openshift_version: "{{ openshift_pkg_version | default('') }}"
  tasks:
  - name: Upgrade to latest available kernel
    action: "{{ ansible_pkg_mgr}} name=kernel state=latest"

  - name: Upgrade master packages
    command: "{{ ansible_pkg_mgr}} update -y {{ openshift.common.service_type }}-master{{ openshift_version }}"

  - name: Ensure python-yaml present for config upgrade
    action: "{{ ansible_pkg_mgr }} name=PyYAML state=present"

  - name: Upgrade master configuration
    openshift_upgrade_config:
      from_version: '3.0'
      to_version: '3.1'
      role: master
      config_base: "{{ hostvars[inventory_hostname].openshift.common.config_base }}"

  - set_fact:
      master_certs_missing: True
      master_cert_subdir: master-{{ openshift.common.hostname }}
      master_cert_config_dir: "{{ openshift.common.config_base }}/master"


- name: Generate missing master certificates
  hosts: oo_first_master
  vars:
    master_hostnames: "{{ hostvars
                          | oo_select_keys(groups.oo_masters_to_config)
                          | oo_collect('openshift.common.all_hostnames')
                          | oo_flatten | unique }}"
    master_generated_certs_dir: "{{ openshift.common.config_base }}/generated-configs"
    masters_needing_certs: "{{ hostvars
                               | oo_select_keys(groups.oo_masters_to_config)
                               | difference([groups.oo_first_master.0]) }}"
    sync_tmpdir: "{{ hostvars.localhost.g_master_mktemp.stdout }}"
    openshift_deployment_type: "{{ deployment_type }}"
  roles:
  - openshift_master_certificates
  post_tasks:
  - name: Remove generated etcd client certs when using external etcd
    file:
      path: "{{ master_generated_certs_dir }}/{{ item.0.master_cert_subdir }}/{{ item.1 }}"
      state: absent
    when: groups.oo_etcd_to_config is defined and groups.oo_etcd_to_config
    with_nested:
    - masters_needing_certs
    - - master.etcd-client.crt
      - master.etcd-client.key

  - name: Create a tarball of the master certs
    command: >
      tar -czvf {{ master_generated_certs_dir }}/{{ item.master_cert_subdir }}.tgz
        -C {{ master_generated_certs_dir }}/{{ item.master_cert_subdir }} .
    with_items: masters_needing_certs

  - name: Retrieve the master cert tarball from the master
    fetch:
      src: "{{ master_generated_certs_dir }}/{{ item.master_cert_subdir }}.tgz"
      dest: "{{ sync_tmpdir }}/"
      flat: yes
      fail_on_missing: yes
      validate_checksum: yes
    with_items: masters_needing_certs


- name: Sync generated certs, update service config and restart master services
  hosts: oo_masters_to_config
  vars:
    sync_tmpdir: "{{ hostvars.localhost.g_master_mktemp.stdout }}"
    openshift_master_ha: "{{ groups.oo_masters_to_config | length > 1 }}"
    openshift_deployment_type: "{{ deployment_type }}"
  tasks:
  - name: Unarchive the tarball on the master
    unarchive:
      src: "{{ sync_tmpdir }}/{{ master_cert_subdir }}.tgz"
      dest: "{{ master_cert_config_dir }}"
    when: inventory_hostname != groups.oo_first_master.0

  - name: Restart master service
    service: name="{{ openshift.common.service_type}}-master" state=restarted
    when: not openshift_master_ha | bool

  - name: Ensure the master service is enabled
    service: name="{{ openshift.common.service_type}}-master" state=started enabled=yes
    when: not openshift_master_ha | bool

  - name: Check for configured cluster
    stat:
      path: /etc/corosync/corosync.conf
    register: corosync_conf
    when: openshift_master_ha | bool

  - name: Destroy cluster
    command: pcs cluster destroy --all
    when: openshift_master_ha | bool and corosync_conf.stat.exists == true
    run_once: true

  - name: Start pcsd
    service: name=pcsd enabled=yes state=started
    when: openshift_master_ha | bool


- name: Re-create cluster
  hosts: oo_first_master
  vars:
    openshift_master_ha: "{{ groups.oo_masters_to_config | length > 1 }}"
    openshift_deployment_type: "{{ deployment_type }}"
    omc_cluster_hosts: "{{ groups.oo_masters_to_config | join(' ') }}"
  roles:
  - role: openshift_master_cluster
    when: openshift_master_ha | bool


- name: Delete temporary directory on localhost
  hosts: localhost
  gather_facts: no
  tasks:
  - file: name={{ g_master_mktemp.stdout }} state=absent
    changed_when: False


- name: Set master update status to complete
  hosts: oo_masters_to_config
  tasks:
  - set_fact:
      master_update_complete: True


##############################################################################
# Gate on master update complete
##############################################################################
- name: Gate on master update
  hosts: localhost
  tasks:
  - set_fact:
      master_update_completed: "{{ hostvars
                                 | oo_select_keys(groups.oo_masters_to_config)
                                 | oo_collect('inventory_hostname', {'master_update_complete': true}) }}"
  - set_fact:
      master_update_failed: "{{ groups.oo_masters_to_config | difference(master_update_completed) }}"
  - fail:
      msg: "Upgrade cannot continue. The following masters did not finish updating: {{ master_update_failed | join(',') }}"
    when: master_update_failed | length > 0


###############################################################################
# Upgrade Nodes
###############################################################################
- name: Upgrade nodes
  hosts: oo_nodes_to_config
  vars:
    openshift_version: "{{ openshift_pkg_version | default('') }}"
  roles:
  - openshift_facts
  tasks:
  - name: Upgrade node packages
    command: "{{ ansible_pkg_mgr }} update -y {{ openshift.common.service_type }}-node{{ openshift_version }}"

  - name: Restart node service
    service: name="{{ openshift.common.service_type }}-node" state=restarted

  - name: Ensure node service enabled
    service: name="{{ openshift.common.service_type }}-node" state=started enabled=yes

  - set_fact:
      node_update_complete: True


##############################################################################
# Gate on nodes update
##############################################################################
- name: Gate on nodes update
  hosts: localhost
  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


###############################################################################
# Post upgrade - Reconcile Cluster Roles and Cluster Role Bindings
###############################################################################
- name: Reconcile Cluster Roles and Cluster Role Bindings
  hosts: oo_masters_to_config
  vars:
    origin_reconcile_bindings: "{{ deployment_type == 'origin' and g_new_version | version_compare('1.0.6', '>') }}"
    ent_reconcile_bindings: true
    openshift_master_ha: "{{ groups.oo_masters_to_config | length > 1 }}"
  tasks:
  - name: Reconcile Cluster Roles
    command: >
      {{ openshift.common.admin_binary}} --config={{ openshift.common.config_base }}/master/admin.kubeconfig
      policy reconcile-cluster-roles --confirm
    run_once: true

  - name: Reconcile Cluster Role Bindings
    command: >
      {{ openshift.common.admin_binary}} --config={{ openshift.common.config_base }}/master/admin.kubeconfig
      policy reconcile-cluster-role-bindings
      --exclude-groups=system:authenticated
      --exclude-groups=system:unauthenticated
      --exclude-users=system:anonymous
      --additive-only=true --confirm
    when: origin_reconcile_bindings | bool or ent_reconcile_bindings | bool
    run_once: true

  - name: Restart master services
    service: name="{{ openshift.common.service_type}}-master" state=restarted
    when: not openshift_master_ha | bool

  - name: Restart master cluster
    command: pcs resource restart master
    when: openshift_master_ha | bool
    run_once: true

  - name: Wait for the clustered master service to be available
    wait_for:
      host: "{{ openshift_master_cluster_vip }}"
      port: 8443
      state: started
      timeout: 180
      delay: 90
    when: openshift_master_ha | bool
    run_once: true

  - set_fact:
      reconcile_complete: True


##############################################################################
# Gate on reconcile
##############################################################################
- name: Gate on reconcile
  hosts: localhost
  tasks:
  - set_fact:
      reconcile_completed: "{{ hostvars
                                 | oo_select_keys(groups.oo_masters_to_config)
                                 | oo_collect('inventory_hostname', {'reconcile_complete': true}) }}"
  - set_fact:
      reconcile_failed: "{{ groups.oo_masters_to_config | difference(reconcile_completed) }}"
  - fail:
      msg: "Upgrade cannot continue. The following masters did not finish reconciling: {{ reconcile_failed | join(',') }}"
    when: reconcile_failed | length > 0




###############################################################################
# Post upgrade - Upgrade default router, default registry and examples
###############################################################################
- name: Upgrade default router and default registry
  hosts: oo_first_master
  vars:
    openshift_deployment_type: "{{ deployment_type }}"
    registry_image: "{{  openshift.master.registry_url | replace( '${component}', 'docker-registry' )  | replace ( '${version}', 'v' + g_new_version  ) }}"
    router_image: "{{ openshift.master.registry_url | replace( '${component}', 'haproxy-router' ) | replace ( '${version}', 'v' + g_new_version ) }}"
    oc_cmd: "{{ openshift.common.client_binary }} --config={{ openshift.common.config_base }}/master/admin.kubeconfig"
  roles:
  # Create the new templates shipped in 3.1, existing templates are left
  # unmodified. This prevents the subsequent role definition for
  # openshift_examples from failing when trying to replace templates that do
  # not already exist. We could have potentially done a replace --force to
  # create and update in one step.
  - openshift_examples
  # Update the existing templates
  - role: openshift_examples
    openshift_examples_import_command: replace
  pre_tasks:
  - name: Check for default router
    command: >
      {{ oc_cmd }} get -n default dc/router
    register: _default_router
    failed_when: false
    changed_when: false

  - name: Check for allowHostNetwork and allowHostPorts
    when: _default_router.rc == 0
    shell: >
      {{ oc_cmd }} get -o yaml scc/privileged | /usr/bin/grep -e allowHostPorts -e allowHostNetwork
    register: _scc

  - name: Grant allowHostNetwork and allowHostPorts
    when:
    - _default_router.rc == 0
    - "'false' in _scc.stdout"
    command: >
      {{ oc_cmd }} patch scc/privileged -p
      '{"allowHostPorts":true,"allowHostNetwork":true}' --api-version=v1

  - name: Update deployment config to 1.0.4/3.0.1 spec
    when: _default_router.rc == 0
    command: >
      {{ oc_cmd }} patch dc/router -p
      '{"spec":{"strategy":{"rollingParams":{"updatePercent":-10},"spec":{"serviceAccount":"router","serviceAccountName":"router"}}}}'
      --api-version=v1

  - name: Switch to hostNetwork=true
    when: _default_router.rc == 0
    command: >
      {{ oc_cmd }} patch dc/router -p '{"spec":{"template":{"spec":{"hostNetwork":true}}}}'
      --api-version=v1

  - name: Update router image to current version
    when: _default_router.rc == 0
    command: >
      {{ oc_cmd }} patch dc/router -p
      '{"spec":{"template":{"spec":{"containers":[{"name":"router","image":"{{ router_image }}"}]}}}}'
      --api-version=v1

  - name: Check for default registry
    command: >
      {{ oc_cmd }} get -n default dc/docker-registry
    register: _default_registry
    failed_when: false
    changed_when: false

  - name: Update registry image to current version
    when: _default_registry.rc == 0
    command: >
      {{ oc_cmd }} patch dc/docker-registry -p
      '{"spec":{"template":{"spec":{"containers":[{"name":"registry","image":"{{ registry_image }}"}]}}}}'
      --api-version=v1