Merge pull request #5672 from ingvagabund/migrate-embedded-etcd

Automatic merge from submit-queue.

Migrate embedded etcd to external etcd

Trello card: https://trello.com/c/9fnBfkT5/528-migrate-embedded-etcd-hosts-to-external-process?menu=filter&filter=label:committed-3.7

Tested on:
- [x] embedded etcd -> external rpm etcd
- [x] embedded etcd -> external docker etcd

One needs to set the `openshift_image_tag` before running the migration to containerized docker etcd.

23 files changed, 167 insertions, 36 deletions

diff --git a/roles/etcd/defaults/main.yaml b/roles/etcd/defaults/main.yaml
index 807b9541a..78f231416 100644
--- a/roles/etcd/defaults/main.yaml
+++ b/roles/etcd/defaults/main.yaml
@@ -70,7 +70,8 @@ etcd_listen_peer_urls: "{{ etcd_peer_url_scheme }}://{{ etcd_ip }}:{{ etcd_peer_
 etcd_advertise_client_urls: "{{ etcd_url_scheme }}://{{ etcd_ip }}:{{ etcd_client_port }}"
 etcd_listen_client_urls: "{{ etcd_url_scheme }}://{{ etcd_ip }}:{{ etcd_client_port }}"
 
-etcd_peer: 127.0.0.1
+# required role variable
+#etcd_peer: 127.0.0.1
 etcdctlv2: "etcdctl --cert-file {{ etcd_peer_cert_file }} --key-file {{ etcd_peer_key_file }} --ca-file {{ etcd_peer_ca_file }} -C https://{{ etcd_peer }}:{{ etcd_client_port }}"
 
 etcd_service: "{{ 'etcd_container' if r_etcd_common_etcd_runtime == 'docker' else 'etcd' }}"
diff --git a/roles/etcd/tasks/auxiliary/clean_data.yml b/roles/etcd/tasks/auxiliary/clean_data.yml
index 95a0e7c0a..1ed2db5bc 100644
--- a/roles/etcd/tasks/auxiliary/clean_data.yml
+++ b/roles/etcd/tasks/auxiliary/clean_data.yml
@@ -1,5 +1,5 @@
 ---
 - name: Remove member data
   file:
-    path: /var/lib/etcd/member
+    path: "{{ etcd_data_dir }}/member"
     state: absent
diff --git a/roles/etcd/tasks/auxiliary/disable_etcd.yml b/roles/etcd/tasks/auxiliary/disable_etcd.yml
new file mode 100644
index 000000000..7c6d0409d
--- /dev/null
+++ b/roles/etcd/tasks/auxiliary/disable_etcd.yml
@@ -0,0 +1,5 @@
+---
+- name: Disable etcd members
+  service:
+    name: "{{ etcd_service }}"
+    state: stopped
diff --git a/roles/etcd/tasks/auxiliary/force_new_cluster.yml b/roles/etcd/tasks/auxiliary/force_new_cluster.yml
new file mode 100644
index 000000000..ae8a36130
--- /dev/null
+++ b/roles/etcd/tasks/auxiliary/force_new_cluster.yml
@@ -0,0 +1,31 @@
+---
+- name: Set ETCD_FORCE_NEW_CLUSTER=true on first etcd host
+  lineinfile:
+    line: "ETCD_FORCE_NEW_CLUSTER=true"
+    dest: /etc/etcd/etcd.conf
+    backup: true
+
+- name: Start etcd
+  systemd:
+    name: "{{ etcd_service }}"
+    state: started
+
+- name: Wait for cluster to become healthy after bringing up first member
+  command: >
+    etcdctl --cert-file {{ etcd_peer_cert_file }} --key-file {{ etcd_peer_key_file }} --ca-file {{ etcd_peer_ca_file }} --endpoint https://{{ etcd_peer }}:{{ etcd_client_port }} cluster-health
+  register: l_etcd_migrate_health
+  until: l_etcd_migrate_health.rc == 0
+  retries: 3
+  delay: 30
+
+- name: Unset ETCD_FORCE_NEW_CLUSTER=true on first etcd host
+  lineinfile:
+    line: "ETCD_FORCE_NEW_CLUSTER=true"
+    dest: /etc/etcd/etcd.conf
+    state: absent
+    backup: true
+
+- name: Restart first etcd host
+  systemd:
+    name: "{{ etcd_service }}"
+    state: restarted
diff --git a/roles/etcd/tasks/backup.archive.yml b/roles/etcd/tasks/backup.archive.yml
new file mode 100644
index 000000000..6daa6dc51
--- /dev/null
+++ b/roles/etcd/tasks/backup.archive.yml
@@ -0,0 +1,3 @@
+---
+- include: backup/vars.yml
+- include: backup/archive.yml
diff --git a/roles/etcd/tasks/backup.copy.yml b/roles/etcd/tasks/backup.copy.yml
new file mode 100644
index 000000000..cc540cbca
--- /dev/null
+++ b/roles/etcd/tasks/backup.copy.yml
@@ -0,0 +1,3 @@
+---
+- include: backup/vars.yml
+- include: backup/copy.yml
diff --git a/roles/etcd/tasks/backup.fetch.yml b/roles/etcd/tasks/backup.fetch.yml
new file mode 100644
index 000000000..26ec15043
--- /dev/null
+++ b/roles/etcd/tasks/backup.fetch.yml
@@ -0,0 +1,3 @@
+---
+- include: backup/vars.yml
+- include: backup/fetch.yml
diff --git a/roles/etcd/tasks/backup.force_new_cluster.yml b/roles/etcd/tasks/backup.force_new_cluster.yml
new file mode 100644
index 000000000..24bd0540d
--- /dev/null
+++ b/roles/etcd/tasks/backup.force_new_cluster.yml
@@ -0,0 +1,12 @@
+---
+- include: backup/vars.yml
+
+- name: Move content of etcd backup under the etcd data directory
+  command: >
+    mv "{{ l_etcd_backup_dir }}/member" "{{ l_etcd_data_dir }}"
+
+- name: Set etcd group for the etcd data directory
+  command: >
+    chown -R etcd:etcd "{{ l_etcd_data_dir }}"
+
+- include: auxiliary/force_new_cluster.yml
diff --git a/roles/etcd/tasks/backup.unarchive.yml b/roles/etcd/tasks/backup.unarchive.yml
new file mode 100644
index 000000000..77a637360
--- /dev/null
+++ b/roles/etcd/tasks/backup.unarchive.yml
@@ -0,0 +1,3 @@
+---
+- include: backup/vars.yml
+- include: backup/unarchive.yml
diff --git a/roles/etcd/tasks/backup/archive.yml b/roles/etcd/tasks/backup/archive.yml
new file mode 100644
index 000000000..f6aa68a6e
--- /dev/null
+++ b/roles/etcd/tasks/backup/archive.yml
@@ -0,0 +1,5 @@
+---
+- name: Archive backup
+  archive:
+    path: "{{ l_etcd_backup_dir }}"
+    dest: "{{ l_etcd_backup_dir }}.tgz"
diff --git a/roles/etcd/tasks/backup/backup.yml b/roles/etcd/tasks/backup/backup.yml
index 42d27c081..ec1a1989c 100644
--- a/roles/etcd/tasks/backup/backup.yml
+++ b/roles/etcd/tasks/backup/backup.yml
@@ -1,21 +1,5 @@
 ---
-# set the etcd backup directory name here in case the tag or sufix consists of dynamic value that changes over time
-# e.g. openshift-backup-{{ lookup('pipe', 'date +%Y%m%d%H%M%S') }} value will change every second so if the date changes
-# right after setting l_etcd_incontainer_backup_dir and before l_etcd_backup_dir facts, the backup directory name is different
-- set_fact:
-    l_backup_dir_name: "openshift-backup-{{ r_etcd_common_backup_tag }}{{ r_etcd_common_backup_sufix_name }}"
-
-- set_fact:
-    l_etcd_data_dir: "{{ etcd_data_dir }}{{ '/etcd.etcd' if r_etcd_common_etcd_runtime == 'runc' else '' }}"
-
-- set_fact:
-    l_etcd_incontainer_data_dir: "{{ etcd_data_dir }}"
-
-- set_fact:
-    l_etcd_incontainer_backup_dir: "{{ l_etcd_incontainer_data_dir }}/{{ l_backup_dir_name }}"
-
-- set_fact:
-    l_etcd_backup_dir: "{{ l_etcd_data_dir }}/{{ l_backup_dir_name }}"
+- include: vars.yml
 
 # TODO: replace shell module with command and update later checks
 - name: Check available disk space for etcd backup
diff --git a/roles/etcd/tasks/backup/copy.yml b/roles/etcd/tasks/backup/copy.yml
new file mode 100644
index 000000000..16604bae8
--- /dev/null
+++ b/roles/etcd/tasks/backup/copy.yml
@@ -0,0 +1,5 @@
+---
+- name: Copy etcd backup
+  copy:
+    src: "{{ etcd_backup_sync_directory }}/{{ l_backup_dir_name }}.tgz"
+    dest: "{{ l_etcd_data_dir }}"
diff --git a/roles/etcd/tasks/backup/fetch.yml b/roles/etcd/tasks/backup/fetch.yml
new file mode 100644
index 000000000..610ce1960
--- /dev/null
+++ b/roles/etcd/tasks/backup/fetch.yml
@@ -0,0 +1,8 @@
+---
+- name: Fetch etcd backup
+  fetch:
+    src: "{{ l_etcd_backup_dir }}.tgz"
+    dest: "{{ etcd_backup_sync_directory }}/"
+    flat: yes
+    fail_on_missing: yes
+    validate_checksum: yes
diff --git a/roles/etcd/tasks/backup/unarchive.yml b/roles/etcd/tasks/backup/unarchive.yml
new file mode 100644
index 000000000..6c75d00a7
--- /dev/null
+++ b/roles/etcd/tasks/backup/unarchive.yml
@@ -0,0 +1,14 @@
+---
+- shell: ls /var/lib/etcd
+  register: output
+
+- debug:
+    msg: "output: {{ output }}"
+
+- name: Unarchive backup
+  # can't use unarchive https://github.com/ansible/ansible/issues/30821
+  # unarchive:
+  #   src: "{{ l_etcd_backup_dir }}.tgz"
+  #   dest: "{{ l_etcd_backup_dir }}"
+  command: >
+    tar -xf "{{ l_etcd_backup_dir }}.tgz" -C "{{ l_etcd_data_dir }}"
diff --git a/roles/etcd/tasks/backup/vars.yml b/roles/etcd/tasks/backup/vars.yml
new file mode 100644
index 000000000..3c009f557
--- /dev/null
+++ b/roles/etcd/tasks/backup/vars.yml
@@ -0,0 +1,18 @@
+---
+# set the etcd backup directory name here in case the tag or sufix consists of dynamic value that changes over time
+# e.g. openshift-backup-{{ lookup('pipe', 'date +%Y%m%d%H%M%S') }} value will change every second so if the date changes
+# right after setting l_etcd_incontainer_backup_dir and before l_etcd_backup_dir facts, the backup directory name is different
+- set_fact:
+    l_backup_dir_name: "openshift-backup-{{ r_etcd_common_backup_tag }}{{ r_etcd_common_backup_sufix_name }}"
+
+- set_fact:
+    l_etcd_data_dir: "{{ etcd_data_dir }}{{ '/etcd.etcd' if r_etcd_common_etcd_runtime == 'runc' else '' }}"
+
+- set_fact:
+    l_etcd_incontainer_data_dir: "{{ etcd_data_dir }}"
+
+- set_fact:
+    l_etcd_incontainer_backup_dir: "{{ l_etcd_incontainer_data_dir }}/{{ l_backup_dir_name }}"
+
+- set_fact:
+    l_etcd_backup_dir: "{{ l_etcd_data_dir }}/{{ l_backup_dir_name }}"
diff --git a/roles/etcd/tasks/backup_master_etcd_certificates.yml b/roles/etcd/tasks/backup_master_etcd_certificates.yml
new file mode 100644
index 000000000..129e1831c
--- /dev/null
+++ b/roles/etcd/tasks/backup_master_etcd_certificates.yml
@@ -0,0 +1,2 @@
+---
+- include: certificates/backup_master_etcd_certificates.yml
diff --git a/roles/etcd/tasks/certificates/backup_master_etcd_certificates.yml b/roles/etcd/tasks/certificates/backup_master_etcd_certificates.yml
new file mode 100644
index 000000000..e65b3e5a2
--- /dev/null
+++ b/roles/etcd/tasks/certificates/backup_master_etcd_certificates.yml
@@ -0,0 +1,7 @@
+---
+- name: Backup master etcd certificates
+  shell: >
+    tar -czvf /etc/origin/master/master-etcd-certificate-backup-{{ ansible_date_time.epoch }}.tgz
+    /etc/origin/master/master.etcd-*
+  args:
+    warn: no
diff --git a/roles/etcd/tasks/check_cluster_health.yml b/roles/etcd/tasks/check_cluster_health.yml
new file mode 100644
index 000000000..75c110972
--- /dev/null
+++ b/roles/etcd/tasks/check_cluster_health.yml
@@ -0,0 +1,2 @@
+---
+- include: migration/check_cluster_health.yml
diff --git a/roles/etcd/tasks/disable_etcd.yml b/roles/etcd/tasks/disable_etcd.yml
new file mode 100644
index 000000000..9202e6e48
--- /dev/null
+++ b/roles/etcd/tasks/disable_etcd.yml
@@ -0,0 +1,2 @@
+---
+- include: auxiliary/disable_etcd.yml
diff --git a/roles/etcd/tasks/fetch_backup.yml b/roles/etcd/tasks/fetch_backup.yml
new file mode 100644
index 000000000..513eed17a
--- /dev/null
+++ b/roles/etcd/tasks/fetch_backup.yml
@@ -0,0 +1,8 @@
+---
+- include: backup/vars.yml
+
+- include: backup/archive.yml
+
+- include: backup/sync_backup.yml
+
+- include: backup/
diff --git a/roles/openshift_master/tasks/check_master_api_is_ready.yml b/roles/openshift_master/tasks/check_master_api_is_ready.yml
new file mode 100644
index 000000000..7e8a7a596
--- /dev/null
+++ b/roles/openshift_master/tasks/check_master_api_is_ready.yml
@@ -0,0 +1,14 @@
+---
+- 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 --silent --tlsv1.2
+    --cacert {{ openshift.common.config_base }}/master/ca-bundle.crt
+    {{ openshift.master.api_url }}/healthz/ready
+  register: l_api_available_output
+  until: l_api_available_output.stdout == 'ok'
+  retries: 120
+  delay: 1
+  run_once: true
+  changed_when: false
diff --git a/roles/openshift_master/tasks/configure_external_etcd.yml b/roles/openshift_master/tasks/configure_external_etcd.yml
new file mode 100644
index 000000000..b0590ac84
--- /dev/null
+++ b/roles/openshift_master/tasks/configure_external_etcd.yml
@@ -0,0 +1,17 @@
+---
+- name: Remove etcdConfig section
+  yedit:
+    src: /etc/origin/master/master-config.yaml
+    key: "etcdConfig"
+    state: absent
+- name: Set etcdClientInfo.ca to master.etcd-ca.crt
+  yedit:
+    src: /etc/origin/master/master-config.yaml
+    key: etcdClientInfo.ca
+    value: master.etcd-ca.crt
+- name: Set etcdClientInfo.urls to the external etcd
+  yedit:
+    src: /etc/origin/master/master-config.yaml
+    key: etcdClientInfo.urls
+    value:
+      - "{{ etcd_peer_url_scheme }}://{{ etcd_ip }}:{{ etcd_peer_port }}"
diff --git a/roles/openshift_master/tasks/main.yml b/roles/openshift_master/tasks/main.yml
index 82b4b420c..824a5886e 100644
--- a/roles/openshift_master/tasks/main.yml
+++ b/roles/openshift_master/tasks/main.yml
@@ -311,23 +311,7 @@
 
 # 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 --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
-  register: l_api_available_output
-  until: l_api_available_output.stdout == 'ok'
-  retries: 120
-  delay: 1
-  run_once: true
-  changed_when: false
+- include: check_master_api_is_ready.yml
   when:
   - openshift.master.cluster_method == 'native'
   - master_api_service_status_changed | bool