Merge "NetApp SolidFire: Fix duplicate volume when API response is lost" into stable/ussuri

cinder/tests/unit/volume/drivers/solidfire/test_solidfire.py

@@ -349,11 +349,14 @@ class SolidFireVolumeTestCase(test.TestCase):
                              'attributes': {'uuid': f_uuid[1]},
                              'qos': None,
                              'iqn': test_name}]}}
-            if params and params['startVolumeID']:
+            if params and params.get('startVolumeID', None):
                 volumes = result['result']['volumes']
-                selected_volumes = [v for v in volumes if v.get('volumeID')
+                selected_volumes = [v for v in volumes if v.get('volumeID') !=
-                                    != params['startVolumeID']]
+                                    params['startVolumeID']]
                 result['result']['volumes'] = selected_volumes
+            else:
+                result = {'result': {'volumes': []}}
+
             return result
         elif method == 'DeleteSnapshot':
             return {'result': {}}

cinder/volume/drivers/solidfire.py

@@ -88,6 +88,7 @@ sf_opts = [
                     '\'usedSpace\', the  driver will report correct '
                     'values as expected by Cinder '
                     'thin provisioning.'),
+
     cfg.IntOpt('sf_api_request_timeout',
                default=30,
                min=30,
@@ -98,7 +99,13 @@ sf_opts = [
                min=60,
                help='Sets time in seconds to wait for a clone of a volume or '
                     'snapshot to complete.'
-               )]
+               ),
+
+    cfg.IntOpt('sf_volume_create_timeout',
+               default=60,
+               min=30,
+               help='Sets time in seconds to wait for a create volume '
+                    'operation to complete.')]
 
 CONF = cfg.CONF
 CONF.register_opts(sf_opts, group=configuration.SHARED_CONF_GROUP)
@@ -241,9 +248,11 @@ class SolidFireDriver(san.SanISCSIDriver):
                     - Implement Active/Active replication support
           2.2.1  - Fix bug #1891914 fix error on cluster workload rebalancing
                    by adding xNotPrimary to the retryable exception list
+          2.2.2  - Fix bug #1896112 SolidFire Driver creates duplicate volume
+                   when API response is lost
     """
 
-    VERSION = '2.2.1'
+    VERSION = '2.2.2'
 
     SUPPORTS_ACTIVE_ACTIVE = True
 
@@ -939,10 +948,62 @@ class SolidFireDriver(san.SanISCSIDriver):
         params['attributes'] = attributes
         return self._issue_api_request('ModifyVolume', params)
 
+    def _list_volumes_by_name(self, sf_volume_name):
+        params = {'volumeName': sf_volume_name}
+        return self._issue_api_request(
+            'ListVolumes', params, version='8.0')['result']['volumes']
+
+    def _wait_volume_is_active(self, sf_volume_name):
+
+        def _wait():
+            volumes = self._list_volumes_by_name(sf_volume_name)
+            if volumes:
+                LOG.debug("Found Volume [%s] in SolidFire backend. "
+                          "Current status is [%s].",
+                          sf_volume_name, volumes[0]['status'])
+                if volumes[0]['status'] == 'active':
+                    raise loopingcall.LoopingCallDone(volumes[0])
+
+        try:
+            timer = loopingcall.FixedIntervalWithTimeoutLoopingCall(
+                _wait)
+            sf_volume = (timer.start(
+                interval=1,
+                timeout=self.configuration.sf_volume_create_timeout).wait())
+
+            return sf_volume
+        except loopingcall.LoopingCallTimeOut:
+            msg = ("Timeout while waiting volume [%s] "
+                   "to be in active state." % sf_volume_name)
+            LOG.error(msg)
+            raise SolidFireAPIException(msg)
+
     def _do_volume_create(self, sf_account, params, endpoint=None):
-        params['accountID'] = sf_account['accountID']
+
-        sf_volid = self._issue_api_request(
+        sf_volume_name = params['name']
-            'CreateVolume', params, endpoint=endpoint)['result']['volumeID']
+        volumes_found = self._list_volumes_by_name(sf_volume_name)
+        if volumes_found:
+            msg = ('Volume name [%s] already exists '
+                   'in SolidFire backend.') % sf_volume_name
+            LOG.error(msg)
+            raise DuplicateSfVolumeNames(message=msg)
+
+        sf_volid = None
+        try:
+            params['accountID'] = sf_account['accountID']
+            response = self._issue_api_request(
+                'CreateVolume', params, endpoint=endpoint)
+            sf_volid = response['result']['volumeID']
+
+        except requests.exceptions.ReadTimeout:
+            LOG.debug("Read Timeout exception caught while creating "
+                      "volume [%s].", sf_volume_name)
+            # Check if volume was created for the given name,
+            # in case the backend has processed the request but failed
+            # to deliver the response before api request timeout.
+            volume_created = self._wait_volume_is_active(sf_volume_name)
+            sf_volid = volume_created['volumeID']
+
         return self._get_model_info(sf_account, sf_volid, endpoint=endpoint)
 
     def _do_snapshot_create(self, params):

releasenotes/notes/sf-fix-duplicate-volume-request-lost-adefacda1298dc62.yaml

@@ -0,0 +1,14 @@
+---
+fixes:
+  - |
+    NetApp SolidFire driver `Bug #1896112
+    <https://bugs.launchpad.net/cinder/+bug/1896112>`_:
+    Fixes an issue that may duplicate volumes during creation, in case
+    the SolidFire backend successfully processes a request and creates
+    the volume, but fails to deliver the result back to the driver (the
+    response is lost). When this scenario occurs, the SolidFire driver
+    will retry the operation, which previously resulted in the creation
+    of a duplicate volume. This fix adds the ``sf_volume_create_timeout``
+    configuration option (default value: 60 seconds) which specifies an
+    additional length of time that the driver will wait for the volume to
+    become active on the backend before raising an exception.