Redistribute failover tests on the modules

- test_failover.py renamed to test_failover_nodes.py
  where are tests for different nodes restart/shutdown
- tests for influxdb moved into separated module
  test_failover_stacklight_services.py
- test 'test_restart_ctl_with_vip' moved from
  test_failover_openstack_services.py to test_failover_nodes.py

This change is required to run failover tests in the separated
jobs.

Change-Id: I4f19a31f4a53107920353e72c2c78f93c9151e10
diff --git a/tcp_tests/tests/system/test_failover_nodes.py b/tcp_tests/tests/system/test_failover_nodes.py
new file mode 100644
index 0000000..0abada9
--- /dev/null
+++ b/tcp_tests/tests/system/test_failover_nodes.py
@@ -0,0 +1,370 @@
+#    Copyright 2017 Mirantis, Inc.
+#
+#    Licensed under the Apache License, Version 2.0 (the "License"); you may
+#    not use this file except in compliance with the License. You may obtain
+#    a copy of the License at
+#
+#         http://www.apache.org/licenses/LICENSE-2.0
+#
+#    Unless required by applicable law or agreed to in writing, software
+#    distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
+#    WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
+#    License for the specific language governing permissions and limitations
+#    under the License.
+import pytest
+
+from tcp_tests import logger
+from tcp_tests.helpers import ext
+
+LOG = logger.logger
+
+
+class TestFailoverNodes(object):
+    """Test class for testing OpenStack nodes failover"""
+
+    @pytest.mark.grab_versions
+    @pytest.mark.fail_snapshot
+    def test_warm_shutdown_ctl01_node(self, underlay, openstack_deployed,
+                                      openstack_actions, show_step):
+        """Test warm shutdown ctl01
+
+        Scenario:
+            1. Prepare salt on hosts
+            2. Setup controller nodes
+            3. Setup compute nodes
+            4. Shutdown ctl01
+            5. Run tempest smoke after failover
+
+
+        """
+        # STEP #1,2,3
+        show_step(1)
+        show_step(2)
+        show_step(3)
+        # STEP #4
+        show_step(4)
+        openstack_actions.warm_shutdown_openstack_nodes('ctl01')
+        # STEP #5
+        show_step(5)
+        openstack_actions.run_tempest(pattern='smoke')
+
+        LOG.info("*************** DONE **************")
+
+    @pytest.mark.grab_versions
+    @pytest.mark.fail_snapshot
+    def test_restart_ctl01_node(self, underlay, openstack_deployed,
+                                openstack_actions, show_step):
+        """Test restart ctl01
+
+        Scenario:
+            1. Prepare salt on hosts
+            2. Setup controller nodes
+            3. Setup compute nodes
+            4. Restart ctl01
+            5. Run tempest smoke after failover
+
+
+        """
+        # STEP #1,2,3
+        show_step(1)
+        show_step(2)
+        show_step(3)
+
+        # STEP #4
+        show_step(4)
+        openstack_actions.warm_restart_nodes('ctl01')
+        # STEP #5
+        show_step(5)
+        openstack_actions.run_tempest(pattern='smoke')
+
+        LOG.info("*************** DONE **************")
+
+    @pytest.mark.grab_versions
+    @pytest.mark.fail_snapshot
+    def test_warm_shutdown_cmp01_node(self, underlay, openstack_deployed,
+                                      openstack_actions, show_step):
+        """Test warm shutdown cmp01
+
+        Scenario:
+            1. Prepare salt on hosts
+            2. Setup controller nodes
+            3. Setup compute nodes
+            4. Shutdown cmp01
+            5. Run tempest smoke after failover
+
+
+        """
+        # STEP #1,2,3
+        show_step(1)
+        show_step(2)
+        show_step(3)
+
+        # STEP #4
+        show_step(4)
+        openstack_actions.warm_shutdown_openstack_nodes('cmp01')
+        # STEP #5
+        show_step(5)
+        openstack_actions.run_tempest(pattern='smoke')
+
+        LOG.info("*************** DONE **************")
+
+    @pytest.mark.grab_versions
+    @pytest.mark.fail_snapshot
+    def test_restart_cmp01_node(self, underlay, openstack_deployed,
+                                openstack_actions, show_step):
+        """Test restart cmp01
+
+        Scenario:
+            1. Prepare salt on hosts
+            2. Setup controller nodes
+            3. Setup compute nodes
+            4. Restart cmp01
+            5. Run tempest smoke after failover
+
+
+        """
+        # STEP #1,2,3
+        show_step(1)
+        show_step(2)
+        show_step(3)
+
+        # STEP #4
+        show_step(4)
+        openstack_actions.warm_restart_nodes('cmp01')
+        # STEP #5
+        show_step(5)
+        openstack_actions.run_tempest(pattern='smoke')
+
+        LOG.info("*************** DONE **************")
+
+    @pytest.mark.grab_versions
+    @pytest.mark.fail_snapshot
+    @pytest.mark.revert_snapshot(ext.SNAPSHOT.sl_deployed)
+    def test_restart_mon01_node(self, openstack_actions,
+                                sl_os_deployed, show_step):
+        """Test restart mon01
+
+        Scenario:
+            1. Prepare salt on hosts
+            2. Setup controller nodes
+            3. Setup compute, monitoring nodes
+            4. Check targets before restart
+            5. Restart mon01
+            6. Check targets after restart
+            7. Run LMA smoke after failover
+
+
+        """
+        # STEP #1,2,3
+        show_step(1)
+        show_step(2)
+        show_step(3)
+
+        # STEP #4
+        show_step(4)
+        mon_nodes = sl_os_deployed.get_monitoring_nodes()
+        LOG.debug('Mon nodes list {0}'.format(mon_nodes))
+        sl_os_deployed.check_prometheus_targets(mon_nodes)
+        before_result = sl_os_deployed.run_sl_tests_json(
+            'cfg01', '/root/stacklight-pytest/stacklight_tests/',
+            'tests/prometheus/', 'test_alerts.py')
+        failed_tests = [test['name'] for test in
+                        before_result if 'passed' not in test['outcome']]
+        # STEP #5
+        show_step(5)
+        openstack_actions.warm_restart_nodes('mon01')
+        # STEP #6
+        show_step(6)
+        sl_os_deployed.check_prometheus_targets(mon_nodes)
+        # STEP #7
+        show_step(7)
+        # Run SL component tetsts
+        after_result = sl_os_deployed.run_sl_tests_json(
+            'cfg01', '/root/stacklight-pytest/stacklight_tests/',
+            'tests/prometheus/', 'test_alerts.py')
+        for test in after_result:
+            if test['name'] not in failed_tests:
+                assert 'passed' in test['outcome'], \
+                    'Failed test {}'.format(test)
+        LOG.info("*************** DONE **************")
+
+    @pytest.mark.grab_versions
+    @pytest.mark.fail_snapshot
+    @pytest.mark.revert_snapshot(ext.SNAPSHOT.sl_deployed)
+    def test_warm_shutdown_mon01_node(self, openstack_actions,
+                                      sl_os_deployed,
+                                      show_step):
+        """Test warm shutdown mon01
+
+        Scenario:
+            1. Prepare salt on hosts
+            2. Setup controller nodes
+            3. Setup compute, monitoring nodes
+            4. Check LMA before mon node shutdown
+            5. Shutdown mon01 node
+            6. Run LMA tests after failover
+
+
+        """
+        # STEP #1,2,3
+        show_step(1)
+        show_step(2)
+        show_step(3)
+
+        # STEP #4
+        show_step(4)
+        mon_nodes = sl_os_deployed.get_monitoring_nodes()
+        LOG.debug('Mon nodes list {0}'.format(mon_nodes))
+        sl_os_deployed.check_prometheus_targets(mon_nodes)
+        before_result = sl_os_deployed.run_sl_tests_json(
+            'cfg01', '/root/stacklight-pytest/stacklight_tests/',
+            'tests/prometheus/', 'test_alerts.py')
+        failed_tests = [test['name'] for test in
+                        before_result if 'passed' not in test['outcome']]
+        # STEP #5
+        show_step(5)
+        openstack_actions.warm_shutdown_openstack_nodes('mon01')
+        # STEP #6
+        show_step(6)
+        # Run SL component tetsts
+        after_result = sl_os_deployed.run_sl_tests_json(
+            'cfg01', '/root/stacklight-pytest/stacklight_tests/',
+            'tests/prometheus/', 'test_alerts.py')
+        for test in after_result:
+            if test['name'] not in failed_tests:
+                assert 'passed' in test['outcome'], \
+                    'Failed test {}'.format(test)
+        LOG.info("*************** DONE **************")
+
+    @pytest.mark.grab_versions
+    @pytest.mark.fail_snapshot
+    @pytest.mark.revert_snapshot(ext.SNAPSHOT.sl_deployed)
+    def test_restart_mon_with_vip(self, sl_os_deployed,
+                                  openstack_actions, salt_actions,
+                                  common_services_actions, show_step):
+        """Test restart mon with VIP
+
+        Scenario:
+            1. Prepare salt on hosts
+            2. Setup controller nodes
+            3. Setup compute and monitoring nodes
+            4. Check LMA before mon node restart
+            5. Find mon minion id with VIP
+            6. Restart mon minion id with VIP
+            7. Check that VIP was actually migrated on a new node
+            8. Run tempest smoke after failover
+
+
+        """
+        # TR case #4753939
+        common_services_actions.check_keepalived_pillar()
+        salt = salt_actions
+
+        # STEP #1,2,3
+        show_step(1)
+        show_step(2)
+        show_step(3)
+
+        # STEP #4
+        show_step(4)
+        mon_nodes = sl_os_deployed.get_monitoring_nodes()
+        LOG.debug('Mon nodes list {0}'.format(mon_nodes))
+        before_result = sl_os_deployed.run_sl_tests_json(
+            'cfg01', '/root/stacklight-pytest/stacklight_tests/',
+            'tests/prometheus/', 'test_alerts.py')
+        failed_tests = [test['name'] for test in
+                        before_result if 'passed' not in test['outcome']]
+
+        # STEP #5
+        show_step(5)
+        mon_vip_pillar = salt.get_pillar(
+            tgt="mon0*",
+            pillar="_param:cluster_vip_address")[0]
+        vip = [vip for minion_id, vip in mon_vip_pillar.items()][0]
+        minion_vip = common_services_actions.get_keepalived_vip_minion_id(vip)
+        LOG.info("VIP {0} is on {1}".format(vip, minion_vip))
+
+        # STEP #6
+        show_step(6)
+        openstack_actions.warm_restart_nodes(minion_vip)
+
+        # STEP #7
+        show_step(7)
+        # Check that VIP has been actually migrated to a new node
+        new_minion_vip = common_services_actions.get_keepalived_vip_minion_id(
+            vip)
+        LOG.info("Migrated VIP {0} is on {1}".format(vip, new_minion_vip))
+        assert new_minion_vip != minion_vip, (
+            "VIP {0} wasn't migrated from {1} after node reboot!"
+            .format(vip, new_minion_vip))
+        common_services_actions.check_keepalived_pillar()
+
+        # STEP #8
+        show_step(8)
+        # Run SL component tetsts
+        after_result = sl_os_deployed.run_sl_tests_json(
+            'cfg01', '/root/stacklight-pytest/stacklight_tests/',
+            'tests/prometheus/', 'test_alerts.py')
+        for test in after_result:
+            if test['name'] not in failed_tests:
+                assert 'passed' in test['outcome'], \
+                    'Failed test {}'.format(test)
+        LOG.info("*************** DONE **************")
+
+    @pytest.mark.grab_versions
+    @pytest.mark.fail_snapshot
+    @pytest.mark.revert_snapshot(ext.SNAPSHOT.openstack_deployed)
+    def test_restart_ctl_with_vip(self, underlay, openstack_deployed,
+                                  openstack_actions, salt_actions,
+                                  common_services_actions, show_step):
+        """Test restart clt with VIP
+
+        Scenario:
+            1. Prepare salt on hosts
+            2. Setup controller nodes
+            3. Setup compute nodes
+            4. Find controller minion id with VIP
+            5. Restart controller minion id with VIP
+            6. Check that VIP was actually migrated on a new node
+            7. Run tempest smoke after failover
+
+
+        """
+        # TR case #3385671
+        common_services_actions.check_keepalived_pillar()
+        salt = salt_actions
+
+        # STEP #1,2,3
+        show_step(1)
+        show_step(2)
+        show_step(3)
+
+        # STEP #4
+        show_step(4)
+        ctl_vip_pillar = salt.get_pillar(
+            tgt="I@nova:controller:enabled:True",
+            pillar="_param:cluster_vip_address")[0]
+        vip = [vip for minion_id, vip in ctl_vip_pillar.items()][0]
+        minion_vip = common_services_actions.get_keepalived_vip_minion_id(vip)
+        LOG.info("VIP {0} is on {1}".format(vip, minion_vip))
+
+        # STEP #5
+        show_step(5)
+        openstack_actions.warm_restart_nodes(minion_vip)
+
+        # STEP #6
+        show_step(6)
+        # Check that VIP has been actually migrated to a new node
+        new_minion_vip = common_services_actions.get_keepalived_vip_minion_id(
+            vip)
+        LOG.info("Migrated VIP {0} is on {1}".format(vip, new_minion_vip))
+        assert new_minion_vip != minion_vip, (
+            "VIP {0} wasn't migrated from {1} after node reboot!"
+            .format(vip, new_minion_vip))
+        common_services_actions.check_keepalived_pillar()
+
+        # STEP #7
+        show_step(7)
+        openstack_actions.run_tempest(pattern='smoke')
+
+        LOG.info("*************** DONE **************")