From e6522bb5a66c159755cd27e92f914514ca1274fd Mon Sep 17 00:00:00 2001
From: Liam Young <liam.young@canonical.com>
Date: Thu, 21 Mar 2019 14:24:02 +0000
Subject: [PATCH] Switch masakari tests to use pacemaker-remote (#203)

Switch masakari tests to use pacemaker-remote. For this the method
of simulating a host failure has to change to sending the process a
sigterm as shuting down pacemaker-remote any other way is seen as an
orderly shutdown and does not trigger things like stonith.

Also add check to ensure all nodes appear to be online from a
crm pov before starting failover test.
---
 zaza/charm_tests/masakari/tests.py |  3 ++
 zaza/configure/masakari.py         | 46 ++++++++++++++++++++++++++++--
 2 files changed, 47 insertions(+), 2 deletions(-)

diff --git a/zaza/charm_tests/masakari/tests.py b/zaza/charm_tests/masakari/tests.py
index 957dbba..36c8d23 100644
--- a/zaza/charm_tests/masakari/tests.py
+++ b/zaza/charm_tests/masakari/tests.py
@@ -26,6 +26,7 @@ import zaza.charm_tests.test_utils as test_utils
 import zaza.utilities.openstack as openstack_utils
 import zaza.utilities.juju as juju_utils
 import zaza.configure.guest
+import zaza.configure.hacluster
 import zaza.configure.masakari
 
 
@@ -111,6 +112,8 @@ class MasakariTest(test_utils.OpenStackBaseTest):
     def test_instance_failover(self):
         """Test masakari managed guest migration."""
         # Launch guest
+        self.assertTrue(zaza.configure.hacluster.check_all_nodes_online(
+            'masakari'))
         vm_name = 'zaza-test-instance-failover'
         self.ensure_guest(vm_name)
 
diff --git a/zaza/configure/masakari.py b/zaza/configure/masakari.py
index 75c6076..fe7de24 100644
--- a/zaza/configure/masakari.py
+++ b/zaza/configure/masakari.py
@@ -129,6 +129,33 @@ def _svc_control(unit_name, action, services, model_name):
         model_name=model_name)
 
 
+def _svc_set_systemd_restart_mode(unit_name, service_name, mode, model_name):
+    """Update the restart mode of the given systemd service.
+
+    :param unit_name: Juju name of unit (app/n)
+    :type unit_name: str
+    :param service_name: Name of systemd service to update
+    :type service_name: str
+    :param mode: Restart mode to switch to eg 'no', 'on-success', 'on-failure',
+                 'on-abort' or 'always'
+    :type mode: str
+    :param model_name: Name of model unit_name resides in.
+    :type model_name: str
+    """
+    # Restart options include: no, on-success, on-failure, on-abort or always
+    logging.info('Setting systemd restart mode for {} to {}'.format(
+        service_name,
+        mode))
+    cmds = [
+        ("sed -i -e 's/^Restart=.*/Restart={}/g' "
+         "/lib/systemd/system/{}.service'").format(mode, service_name),
+        'systemctl daemon-reload']
+    logging.info('Running {} on {}'.format(cmds, unit_name))
+    zaza.model.run_on_unit(
+        unit_name, command=';'.join(cmds),
+        model_name=model_name)
+
+
 def simulate_compute_host_failure(unit_name, model_name):
     """Simulate compute node failure from a masakari and nova POV.
 
@@ -142,11 +169,21 @@ def simulate_compute_host_failure(unit_name, model_name):
     :type model_name: str
     """
     logging.info('Simulating failure of compute node {}'.format(unit_name))
+    _svc_set_systemd_restart_mode(
+        unit_name,
+        'pacemaker_remote',
+        'no',
+        model_name)
     _svc_control(
         unit_name,
         'stop',
-        ['corosync', 'pacemaker', 'nova-compute'],
+        ['corosync', 'nova-compute'],
         model_name)
+    logging.info('Sending pacemaker_remoted a SIGTERM')
+    zaza.model.run_on_unit(
+        unit_name,
+        'pkill -9 -f /usr/sbin/pacemaker_remoted',
+        model_name=model_name)
 
 
 def simulate_compute_host_recovery(unit_name, model_name):
@@ -162,10 +199,15 @@ def simulate_compute_host_recovery(unit_name, model_name):
     :type model_name: str
     """
     logging.info('Simulating recovery of compute node {}'.format(unit_name))
+    _svc_set_systemd_restart_mode(
+        unit_name,
+        'pacemaker_remote',
+        'on-failure',
+        model_name)
     _svc_control(
         unit_name,
         'start',
-        ['corosync', 'pacemaker', 'nova-compute'],
+        ['corosync', 'pacemaker_remote', 'nova-compute'],
         model_name)