okd/roles/openshift_logging_elasticsearch/tasks/rolling_cluster_restart.yml

---
# If we are currently restarting the "es" cluster we want to check if we are scaling up the number of es nodes
# If we are currently restarting the "es-ops" cluster we want to check if we are scaling up the number of ops nodes
# If we've created a new node for that cluster then the appropriate variable will be true, otherwise we default to false
- set_fact:
    _skip_healthcheck: "{{ ( __logging_scale_up | default(false) ) if _cluster_component == 'es' else ( __logging_ops_scale_up | default(false) ) }}"

# Flush ES
# It is possible for this to fail on a brand new cluster, so don't fail then
- name: "Flushing for logging-{{ _cluster_component }} cluster"
  command: >
    curl -s -k
    --cert {{ _logging_handler_tempdir.stdout }}/admin-cert
    --key {{ _logging_handler_tempdir.stdout }}/admin-key
    -XPOST 'https://logging-{{ _cluster_component }}.{{ openshift_logging_elasticsearch_namespace }}.svc:9200/_flush/synced'
  register: _flush_output
  changed_when:
  - "_flush_output.stdout != ''"
  - (_flush_output.stdout | from_json)['_shards']['successful'] > 0
  failed_when: false

# if we are skipping the health check, then we should only disable and enable shard allocation once for the cluster
- when: _skip_healthcheck | bool
  name: "Disable shard balancing for logging-{{ _cluster_component }} cluster"
  command: >
    curl -s -k
    --cert {{ _logging_handler_tempdir.stdout }}/admin-cert
    --key {{ _logging_handler_tempdir.stdout }}/admin-key
    -XPUT 'https://logging-{{ _cluster_component }}.{{ openshift_logging_elasticsearch_namespace }}.svc:9200/_cluster/settings'
    -d '{ "transient": { "cluster.routing.allocation.enable" : "none" } }'
  register: _cluster_disable_output
  changed_when:
  - "_cluster_disable_output.stdout != ''"
  - (_cluster_disable_output.stdout | from_json)['acknowledged'] | bool
  failed_when: false

# Loop over each DC for restart_es_node.yml
- include_tasks: restart_es_node.yml
  with_items: "{{ logging_restart_cluster_dcs }}"
  loop_control:
    loop_var: _es_node

# if we are skipping the health check, then we should only disable and enable shard allocation once for the cluster
- when:
  - _skip_healthcheck | bool
  - "_cluster_disable_output.stdout != ''"
  - (_cluster_disable_output.stdout | from_json)['acknowledged'] | bool
  name: "Waiting for ES cluster logging{{ _cluster_component }} to be up"
  command: >
    curl -s -k
    --cert {{ _logging_handler_tempdir.stdout }}/admin-cert
    --key {{ _logging_handler_tempdir.stdout }}/admin-key
    --max-time 30
    -o /dev/null \
    -w '%{response_code}'
    https://logging-{{ _cluster_component }}.{{ openshift_logging_elasticsearch_namespace }}.svc:9200/
  register: _cluster_status
  until: "_cluster_status.stdout == '200'"
  retries: "{{ __elasticsearch_ready_retries }}"
  delay: 30
  changed_when: false
  failed_when: false

- when:
  - _skip_healthcheck | bool
  - "_cluster_disable_output.stdout != ''"
  - (_cluster_disable_output.stdout | from_json)['acknowledged'] | bool
  name: "Enable shard balancing for logging-{{ _cluster_component }} cluster"
  command: >
    curl -s -k
    --cert {{ _logging_handler_tempdir.stdout }}/admin-cert
    --key {{ _logging_handler_tempdir.stdout }}/admin-key
    -XPUT 'https://logging-{{ _cluster_component }}.{{ openshift_logging_elasticsearch_namespace }}.svc:9200/_cluster/settings'
    -d '{ "transient": { "cluster.routing.allocation.enable" : "all" } }'
  register: _cluster_enable_output
  changed_when:
  - "_cluster_enable_output.stdout != ''"
  - (_cluster_enable_output.stdout | from_json)['acknowledged'] | bool