[SVF]: Fixed host and group failback issues

[Spectrum Virtualize Family] Fixed issue regarding failback of a
cinder host and a group that impacts IBM Spectrum Virtualize Family
storage back-end performance and applies to mirror replication types
such as metro, global and GMCV volume replication types.

closes bug: #1906528

Change-Id: I43ff6533436bb8467862d9b6e9dd3fc82290d20d

cinder/tests/unit/volume/drivers/ibm/test_storwize_svc.py

@@ -11505,10 +11505,6 @@ class StorwizeSVCReplicationTestCase(test.TestCase):
                           self.driver.failover_replication, self.ctxt, group,
                           vols, self.fake_target['backend_id'])
 
-        self.assertRaises(exception.UnableToFailOver,
-                          self.driver.failover_replication, self.ctxt, group,
-                          vols, storwize_const.FAILBACK_VALUE)
-
         with mock.patch.object(storwize_svc_common.StorwizeSSH,
                                'stoprcconsistgrp') as stoprccg:
             stoprccg.side_effect = exception.VolumeBackendAPIException(

cinder/volume/drivers/ibm/storwize_svc/replication.py

@@ -56,8 +56,21 @@ class StorwizeSVCReplication(object):
             LOG.exception(msg)
             raise exception.VolumeDriverException(message=msg)
 
+    @volume_utils.trace
     def replication_failback(self, volume):
-        pass
+        tgt_volume = storwize_const.REPLICA_AUX_VOL_PREFIX + volume['name']
+        rel_info = self.target_helpers.get_relationship_info(tgt_volume)
+        if rel_info:
+            try:
+                self.target_helpers.stop_relationship(tgt_volume, access=True)
+                self.target_helpers.start_relationship(tgt_volume, 'master')
+                return
+            except Exception as e:
+                msg = (_('Unable to fail-back the volume: %(vol)s to the '
+                         'master back-end, error: %(error)s') %
+                       {"vol": volume['name'], "error": six.text_type(e)})
+                LOG.exception(msg)
+                raise exception.VolumeDriverException(message=msg)
 
     def volume_replication_setup(self, context, vref):
         pass
@@ -108,21 +121,6 @@ class StorwizeSVCReplicationGlobalMirror(StorwizeSVCReplication):
             raise exception.VolumeDriverException(message=msg)
         LOG.debug('leave: volume_replication_setup:volume %s', vref['name'])
 
-    def replication_failback(self, volume):
-        tgt_volume = storwize_const.REPLICA_AUX_VOL_PREFIX + volume['name']
-        rel_info = self.target_helpers.get_relationship_info(tgt_volume)
-        if rel_info:
-            try:
-                self.target_helpers.switch_relationship(rel_info['name'],
-                                                        aux=False)
-                return
-            except Exception as e:
-                msg = (_('Unable to fail-back the volume:%(vol)s to the '
-                         'master back-end, error:%(error)s') %
-                       {"vol": volume['name'], "error": e})
-                LOG.exception(msg)
-                raise exception.VolumeDriverException(message=msg)
-
 
 class StorwizeSVCReplicationMetroMirror(
         StorwizeSVCReplicationGlobalMirror):
@@ -232,23 +230,6 @@ class StorwizeSVCReplicationGMCV(StorwizeSVCReplicationGlobalMirror):
             raise exception.VolumeDriverException(message=msg)
         LOG.debug('leave: volume_replication_setup:volume %s', vref['name'])
 
-    def replication_failback(self, volume):
-        LOG.debug('enter: replication_failback: volume=%(volume)s',
-                  {'volume': volume['name']})
-        tgt_volume = storwize_const.REPLICA_AUX_VOL_PREFIX + volume['name']
-        rel_info = self.target_helpers.get_relationship_info(tgt_volume)
-        if rel_info:
-            try:
-                self.target_helpers.stop_relationship(tgt_volume, access=True)
-                self.target_helpers.start_relationship(tgt_volume, 'master')
-                return
-            except Exception as e:
-                msg = (_('Unable to fail-back the volume:%(vol)s to the '
-                         'master back-end, error:%(error)s') %
-                       {"vol": volume['name'], "error": six.text_type(e)})
-                LOG.exception(msg)
-                raise exception.VolumeDriverException(message=msg)
-
 
 class StorwizeSVCReplicationManager(object):
 

cinder/volume/drivers/ibm/storwize_svc/storwize_svc_common.py

@@ -4295,6 +4295,7 @@ class StorwizeSVCCommonDriver(san.SanDriver,
             volumes_model_update.append(volume_model_update)
         return model_update, volumes_model_update
 
+    @volume_utils.trace
     def _rep_grp_failback(self, ctxt, group, sync_grp=True):
         """Fail back all the volume in the replication group."""
         model_update = {
@@ -4302,59 +4303,14 @@ class StorwizeSVCCommonDriver(san.SanDriver,
         rccg_name = self._get_rccg_name(group)
 
         try:
-            self._master_backend_helpers.get_system_info()
-        except Exception as ex:
-            msg = (_("Unable to failback group %(rccg)s due to primary is not "
-                     "reachable. error=%(error)s"),
-                   {'rccg': rccg_name, 'error': ex})
-            LOG.error(msg)
-            raise exception.UnableToFailOver(reason=msg)
-
-        rccg = self._helpers.get_rccg(rccg_name)
-        if not rccg:
-            msg = (_("Unable to failback group %(rccg)s due to replication "
-                     "group does not exist on backend."),
-                   {'rccg': rccg_name})
-            LOG.error(msg)
-            raise exception.UnableToFailOver(reason=msg)
-
-        if rccg['relationship_count'] == '0':
-            msg = (_("Unable to failback empty group %(rccg)s"),
-                   {'rccg': rccg['name']})
-            LOG.error(msg)
-            raise exception.UnableToFailOver(reason=msg)
-
-        if rccg['primary'] == 'master':
-            LOG.info("Do not need to fail back group %(rccg)s again due to "
-                     "primary is already master.", {'rccg': rccg['name']})
+            self._aux_backend_helpers.stop_rccg(rccg_name, access=True)
+            self._aux_backend_helpers.start_rccg(rccg_name, primary='master')
             return model_update
-
-        if sync_grp:
-            self._sync_with_aux_grp(ctxt, rccg['name'])
-            self._wait_replica_grp_ready(ctxt, rccg['name'])
-
-        if rccg['cycling_mode'] == 'multi':
-            # This is a gmcv replication group
-            try:
-                self._aux_backend_helpers.stop_rccg(rccg['name'], access=True)
-                self._aux_backend_helpers.start_rccg(rccg['name'],
-                                                     primary='master')
-                return model_update
-            except exception.VolumeBackendAPIException as e:
-                msg = (_('Unable to fail over the group %(rccg)s to the aux '
-                         'back-end, error: %(error)s') %
-                       {"rccg": rccg['name'], "error": e})
-                LOG.exception(msg)
-                raise exception.UnableToFailOver(reason=msg)
-        else:
-            try:
-                self._helpers.switch_rccg(rccg['name'], aux=False)
-            except exception.VolumeBackendAPIException as e:
-                msg = (_('Unable to fail back the group %(rccg)s, error: '
-                         '%(error)s') % {"rccg": rccg['name'], "error": e})
-                LOG.exception(msg)
-                raise exception.UnableToFailOver(reason=msg)
-        return model_update
+        except exception.VolumeBackendAPIException as e:
+            msg = (_('Unable to fail back the group %(rccg)s, error: '
+                     '%(error)s') % {"rccg": rccg_name, "error": e})
+            LOG.exception(msg)
+            raise exception.UnableToFailOver(reason=msg)
 
     @volume_utils.trace
     def _rep_grp_failover(self, ctxt, group):

releasenotes/notes/bug-1906528-ibm-svf-fix-host-failback-switch-relationship-b5b7320811688cda.yaml

@@ -0,0 +1,7 @@
+---
+fixes:
+  - |
+    `Bug #1906528 <https://bugs.launchpad.net/cinder/+bug/1906528>`_:
+    IBM Spectrum Virtualize Family driver: Fixed issue regarding
+    host-failback and group-failback which impacts storage back-end
+    performance.