double failover test using postgres-maintenance role, refs #33389

.gitlab-ci.yml

@@ -103,7 +103,7 @@ stop-ovh-preprod:
   script:
     - make stop-preprod-ovh
 
-test-ha-pgsql:
+test-pgsql-ha:
   image: registry.ubicast.net/mediaserver/envsetup:root
   stage: test-pgsql-ha
   tags:
@@ -115,7 +115,7 @@ test-ha-pgsql:
       changes:
         - ansible/**/*
   script:
-    - make test ha-pgsql=1
+    - make test pgsql-ha=1
 
 test-mediaserver-ha:
   image: registry.ubicast.net/mediaserver/envsetup:root

Makefile

@@ -18,8 +18,11 @@ endif
 ifdef keep
 	MOLECULE_TEST_FLAGS += --destroy=never --parallel
 endif
-ifdef ha-pgsql
-	MOLECULE_TEST_FLAGS += --scenario-name ha-pgsql
+ifdef pgsql-ha
+	MOLECULE_TEST_FLAGS += --scenario-name pgsql-ha
+endif
+ifdef mediaserver-ha
+	MOLECULE_TEST_FLAGS += --scenario-name mediaserver-ha
 endif
 ifdef mediaserver-ha
 	MOLECULE_TEST_FLAGS += --scenario-name mediaserver-ha
@@ -65,7 +68,7 @@ lint:
 	ANSIBLE_CONFIG=$(ANSIBLE_CONFIG) $(ANSIBLE_LINT_BIN) ansible/playbooks/site.yml
 
 .PHONY: test
-## test: Run development tests on the project : debug=1, keep=1, SKYREACH_SYSTEM_KEY=<xxx>, ha-pgsql=1
+## test: Run development tests on the project : debug=1, keep=1, SKYREACH_SYSTEM_KEY=<xxx>, pgsql-ha=1
 test:
 	cd ansible; molecule $(MOLECULE_FLAGS) test $(MOLECULE_TEST_FLAGS)
 

ansible/molecule/ha-pgsql/tests/test_postgres_e_shutdown_primary.py

-import os
-
-import testinfra.utils.ansible_runner
-
-import time
-
-
-hosts = testinfra.utils.ansible_runner.AnsibleRunner(os.environ["MOLECULE_INVENTORY_FILE"]).get_hosts("postgres")
-testinfra_hosts = [i for i in hosts if i.startswith('db0')]
-
-
-def test_postgresql_check_shutdown_primary(host):
-    ''' Shutdown the primary server '''
-
-    s = host.ansible("command", "systemctl stop postgresql", become=True, check=False)
-    assert s['changed']
-
-    time.sleep(40)
-
-    s = host.socket("tcp://127.0.0.1:5432")
-    assert not s.is_listening

ansible/molecule/ha-pgsql/tests/test_postgres_f_new_primary_cluster_status.py

-import os
-
-import testinfra.utils.ansible_runner
-
-import commons
-
-# /!\ This test run accross all servers
-testinfra_hosts = testinfra.utils.ansible_runner.AnsibleRunner(os.environ["MOLECULE_INVENTORY_FILE"]).get_hosts("postgres")
-
-
-def test_postgresql_check_repmgr_new_master(host):
-    ''' check repmgr status for each node after new master election '''
-
-    if host.ansible.get_variables()["inventory_hostname"].startswith("db0-default"):
-        data = commons.get_status(host)
-        assert data == "fenced"
-    if host.ansible.get_variables()["inventory_hostname"].startswith("db1-default"):
-        data = commons.get_status(host)
-        assert data == "primary"
-    if host.ansible.get_variables()["inventory_hostname"].startswith("db2-default"):
-        data = commons.get_status(host)
-        assert data == "witness"

ansible/molecule/ha-pgsql/tests/test_postgres_g_new_primary_write.py

-import os
-
-import testinfra.utils.ansible_runner
-
-
-hosts = testinfra.utils.ansible_runner.AnsibleRunner(os.environ["MOLECULE_INVENTORY_FILE"]).get_hosts("postgres")
-testinfra_hosts = [i for i in hosts if i.startswith('db1')]
-
-
-def test_postgresql_insert_new_master(host):
-    s = host.ansible("postgresql_query", "db=test query='INSERT INTO test_ha (name) VALUES (\'test2\');'", become=True, check=False, become_user='postgres')
-    assert s["changed"]

ansible/molecule/ha-pgsql/tests/test_postgres_h_reintegrate_server.py

-import os
-
-import testinfra.utils.ansible_runner
-import time
-
-hosts = testinfra.utils.ansible_runner.AnsibleRunner(os.environ["MOLECULE_INVENTORY_FILE"]).get_hosts("postgres")
-
-
-testinfra_hosts = [i for i in hosts if i.startswith('db0')]
-
-
-def test_postgresql_delete_data(host):
-    ''' delete data directory '''
-
-    s = host.ansible("command", "rm -rf /var/lib/postgresql/11/main/", become=True, check=False)
-    assert s['changed']
-
-
-def test_postgresql_launch_repmgr_sync(host):
-    ''' sync data with primary server using repmgr '''
-
-    current_m = [i for i in hosts if i.startswith('db1')]
-    # print("current master" + current_m[0])
-    current_master = testinfra.utils.ansible_runner.AnsibleRunner(os.environ["MOLECULE_INVENTORY_FILE"]).get_host(current_m[0])
-    # current_master = testinfra.utils.ansible_runner.AnsibleRunner(os.environ["MOLECULE_INVENTORY_FILE"]).get_host("db1-default")
-    # print(current_master)
-    current_master_ip = current_master.interface('eth0').addresses[0]
-    rep_mgr_command = "repmgr -f /etc/postgresql/11/main/repmgr.conf --force --verbose standby clone -h " + str(current_master_ip) + " -d repmgr -U repmgr -c"
-    s = host.ansible("command", rep_mgr_command, become=True, become_user='postgres', check=False)
-    assert s['changed']
-
-
-def test_postgresql_start_postgresql(host):
-    ''' start postgresql '''
-
-    s = host.ansible("command", "systemctl start postgresql", become=True, check=False)
-    time.sleep(20)
-    assert s['changed']
-
-
-def test_pogresql_register_as_standby(host):
-    ''' register server as standby in repmgr '''
-
-    s = host.ansible("command", "repmgr -f /etc/postgresql/11/main/repmgr.conf --force --verbose standby register", become=True, become_user='postgres', check=False)
-    time.sleep(20)
-    assert s['changed']

ansible/molecule/ha-pgsql/tests/test_postgres_i_final_cluster_status.py

-import os
-
-import testinfra.utils.ansible_runner
-
-import commons
-
-# /!\ This test run accross all servers
-testinfra_hosts = testinfra.utils.ansible_runner.AnsibleRunner(os.environ["MOLECULE_INVENTORY_FILE"]).get_hosts("postgres")
-
-
-def test_postgresql_check_status_after_shutdown(host):
-    ''' check repmgr status accross server after primary change and server reintegration '''
-
-    if host.ansible.get_variables()["inventory_hostname"].startswith("db0"):
-        data = commons.get_status(host)
-        assert data == "standby"
-    if host.ansible.get_variables()["inventory_hostname"].startswith("db1"):
-        data = commons.get_status(host)
-        assert data == "primary"
-    if host.ansible.get_variables()["inventory_hostname"].startswith("db2"):
-        data = commons.get_status(host)
-        assert data == "witness"

ansible/molecule/ha-pgsql/molecule.yml → ansible/molecule/pgsql-ha/molecule.yml

@@ -49,6 +49,8 @@ provisioner:
     ANSIBLE_ACTION_PLUGINS: ../../plugins/action
     ANSIBLE_PYTHON_INTERPRETER: /usr/bin/python3
     SKYREACH_SYSTEM_KEY: s1121eb6e7593525bf3e0302586c82d2
+  playbooks:
+    side_effect: side_effect.yml
 verifier:
   name: testinfra
   # options:

ansible/molecule/pgsql-ha/side_effect.yml

+---
+- name: first failover
+  import_playbook: side_effect_destroy_master.yml
+
+# Refresh inventory to have a clean inventory (each server in the initial group)
+- name: refresh inventory
+  hosts: all
+  tasks:
+  - name: refresh inventory
+    meta: refresh_inventory
+
+- name: second failover
+  import_playbook: side_effect_destroy_master.yml

ansible/molecule/pgsql-ha/side_effect_destroy_master.yml

+---
+- name: stop master node
+  hosts: postgres[0]
+  tasks:
+  - name: stop postgresql
+    service:
+      name: postgresql@11-main
+      state: stopped
+
+- name: pause for 45 seconds
+  hosts: localhost
+  tasks:
+    - pause:
+        seconds: 45
+
+- name: set correct host to correct groups
+  hosts: postgres
+  tasks:
+    - name: add db0 to fenced group
+      add_host:
+        name: "{{ groups['postgres'][0] }}"
+        groups: [ 'postgres_fenced', 'postgres_standby' ]
+
+    - name: add db1 to primary group
+      add_host:
+        name: "{{ groups['postgres'][1] }}"
+        groups: postgres_primary
+
+# db0 is now fenced, import playbook to switch it back to standby
+- import_playbook: ../../playbooks/postgres-maintenance/rephacheck_status.yml
+- import_playbook: ../../playbooks/postgres-maintenance/fenced_to_standby.yml
+
+- name: pause for 45 seconds
+  hosts: localhost
+  tasks:
+    - pause:
+        seconds: 45
+
+# db0 is now standby, import playbook to switch it back as primary
+- import_playbook: ../../playbooks/postgres-maintenance/rephacheck_status.yml
+- import_playbook: ../../playbooks/postgres-maintenance/standby_to_primary.yml
+
+- name: pause for 45 seconds
+  hosts: localhost
+  tasks:
+    - pause:
+        seconds: 45
+
+# db0 is now primary, restart repmgrd
+- import_playbook: ../../playbooks/postgres-maintenance/rephacheck_status.yml
+- import_playbook: ../../playbooks/postgres-maintenance/restart_repmgrd.yml
+
+...

ansible/playbooks/postgres-maintenance.yml

 #!/usr/bin/env ansible-playbook
 ---
 
-- name: GATHER FACTS
-  hosts: postgres_primary:postgres_standby:postgres_fenced
-  tags: always
-  tasks:
-    - name: get cluster state
-      command: "rephacheck"
-      register: rephacheck
-    - name: show status for each node
-      debug:
-        msg: "Current node {{ ansible_hostname }} status {{ rephacheck['stdout'] }}"
-      when: rephacheck['stdout'] != ""
+- import_playbook: postgres-maintenance/rephacheck_status.yml
+  tags: [ 'always' ]
 
-- name: POSTGRESQL SWITCH CURRENT STANDBY TO PRIMARY
-  hosts: postgres_standby
-  tags: [ 'never', 'standby-to-primary' ]
-  tasks:
-    - name: fail if node status if not standby
-      fail:
-        msg: "Current status {{ rephacheck['stdout'] }} must be standby."
-      when: rephacheck['stdout'] != "standby"
-    - name: check if node is currently in standby
-      command: "repmgr standby switchover -f /etc/postgresql/11/main/repmgr.conf --siblings-follow --dry-run"
-      become: true
-      become_user: postgres
-      when: rephacheck['stdout'] == "standby"
-      register: standby_dry_run
-    - name: switch standby node to primary
-      command: "repmgr standby switchover -f /etc/postgresql/11/main/repmgr.conf --siblings-follow"
-      become: true
-      become_user: postgres
-      when:
-        - standby_dry_run is succeeded
-        - rephacheck['stdout'] == "standby"
-
-- name: POSTGRESQL SWITCH CURRENT FENCED TO STANDBY
-  hosts: postgres_fenced
+- import_playbook: postgres-maintenance/fenced_to_standby.yml
   tags: [ 'never', 'fenced-to-standby' ]
-  tasks:
-    - name: fail if node status if not fenced
-      fail:
-        msg: "Current status {{ rephacheck['stdout'] }} must be fenced."
-      when: rephacheck['stdout'] != "fenced"
-    - name: stop postgresql
-      systemd:
-        name: postgresql
-        state: stopped
-    - name: delete postgresql data directory
-      file:
-        path: /var/lib/postgresql/11/main/
-        state: absent
-        force: true
-    - name: copy data from primary
-      command: "repmgr -f /etc/postgresql/11/main/repmgr.conf --force --verbose standby clone -h {{ hostvars[groups['postgres_primary'][0]]['ansible_default_ipv4']['address'] }} -d repmgr -U repmgr -c"
-      become: true
-      become_user: postgres
-      register: copy_from_primary
-    - name: start postgresql
-      systemd:
-        name: postgresql
-        state: started
-      when: copy_from_primary is succeeded
-    - name: register node as standby
-      command: "repmgr -f /etc/postgresql/11/main/repmgr.conf --force --verbose standby register"
-      become: true
-      become_user: postgres
-      when: copy_from_primary is succeeded
 
-- name: REPMGRD RESTART
-  hosts: postgres
+- import_playbook: postgres-maintenance/standby_to_primary.yml
+  tags: [ 'never', 'standby-to-primary' ]
+
+- import_playbook: postgres-maintenance/restart_repmgrd.yml
   tags: [ 'never', 'restart-repmgrd', 'standby-to-primary' ]
-  tasks:
-    - name: kill repmgrd
-      command: "pkill repmgrd"
-      ignore_errors: true
-    - name: restart repmgrd
-      systemd:
-        name: repmgrd
-        state: restarted
+
 ...

ansible/playbooks/postgres-maintenance/fenced_to_standby.yml

+#!/usr/bin/env ansible-playbook
+---
+- name: POSTGRESQL SWITCH CURRENT FENCED TO STANDBY
+  hosts: postgres_fenced
+  tasks:
+    - name: fail if node status if not fenced
+      fail:
+        msg: "Current status {{ rephacheck['stdout'] }} must be fenced."
+      when: rephacheck['stdout'] != "fenced"
+    - name: stop postgresql
+      systemd:
+        name: postgresql
+        state: stopped
+    - name: delete postgresql data directory
+      file:
+        path: /var/lib/postgresql/11/main/
+        state: absent
+        force: true
+    - name: copy data from primary
+      command: "repmgr -f /etc/postgresql/11/main/repmgr.conf --force --verbose standby clone -h {{ hostvars[groups['postgres_primary'][0]]['ansible_default_ipv4']['address'] }} -d repmgr -U repmgr -c"
+      become: true
+      become_user: postgres
+      register: copy_from_primary
+    - name: start postgresql
+      systemd:
+        name: postgresql
+        state: started
+      when: copy_from_primary is succeeded
+    - name: register node as standby
+      command: "repmgr -f /etc/postgresql/11/main/repmgr.conf --force --verbose standby register"
+      become: true
+      become_user: postgres
+      when: copy_from_primary is succeeded
+
+...

ansible/playbooks/postgres-maintenance/rephacheck_status.yml

+#!/usr/bin/env ansible-playbook
+---
+- name: REHACHECK STATUS
+  hosts: postgres_primary:postgres_standby:postgres_fenced
+  tasks:
+    - name: get cluster state
+      command: "rephacheck"
+      register: rephacheck
+    - name: show status for each node
+      debug:
+        msg: "Current node {{ ansible_hostname }} status {{ rephacheck['stdout'] }}"
+      when: rephacheck['stdout'] != ""
+
+...

ansible/playbooks/postgres-maintenance/restart_repmgrd.yml

+#!/usr/bin/env ansible-playbook
+---
+- name: REPMGRD RESTART
+  hosts: postgres
+  tasks:
+    - name: kill repmgrd
+      command: "pkill repmgrd"
+      ignore_errors: true
+    - name: restart repmgrd
+      systemd:
+        name: repmgrd
+        state: restarted
+
+...