diff --git a/nova/compute/manager.py b/nova/compute/manager.py
index ab6b5dcf6d2a..253083ca8427 100644
--- a/nova/compute/manager.py
+++ b/nova/compute/manager.py
@@ -1311,13 +1311,8 @@ class ComputeManager(manager.Manager):
         # TODO(yamahata): eliminate dumb polling
         start = time.time()
         retries = CONF.block_device_allocate_retries
-        if retries < 0:
-            LOG.warning("Treating negative config value (%(retries)s) for "
-                        "'block_device_retries' as 0.",
-                        {'retries': retries})
-        # (1) treat  negative config value as 0
-        # (2) the configured value is 0, one attempt should be made
-        # (3) the configured value is > 0, then the total number attempts
+        # (1) if the configured value is 0, one attempt should be made
+        # (2) if the configured value is > 0, then the total number attempts
         #      is (retries + 1)
         attempts = 1
         if retries >= 1:
diff --git a/nova/conf/compute.py b/nova/conf/compute.py
index 159024306bac..a7d68dea7372 100644
--- a/nova/conf/compute.py
+++ b/nova/conf/compute.py
@@ -639,6 +639,7 @@ Possible values:
 """),
     cfg.IntOpt('block_device_allocate_retries',
         default=60,
+        min=0,
         help="""
 The number of times to check for a volume to be "available" before attaching
 it during server create.
@@ -662,7 +663,6 @@ Possible values:
 
 * 60 (default)
 * If value is 0, then one attempt is made.
-* Any negative value is treated as 0.
 * For any value > 0, total attempts are (value + 1)
 
 Related options:
diff --git a/nova/tests/unit/compute/test_compute.py b/nova/tests/unit/compute/test_compute.py
index a55a87575a90..c5c9dd4013c9 100644
--- a/nova/tests/unit/compute/test_compute.py
+++ b/nova/tests/unit/compute/test_compute.py
@@ -636,21 +636,6 @@ class ComputeVolumeTestCase(BaseTestCase):
         attempts = c._await_block_device_map_created(self.context, '1')
         self.assertEqual(attempts, 3)
 
-    def test_await_block_device_created_retries_negative(self):
-        c = self.compute
-        self.flags(block_device_allocate_retries=-1)
-        self.flags(block_device_allocate_retries_interval=0.1)
-
-        def volume_get(self, context, vol_id):
-            return {
-                'status': 'available',
-                'id': 'blah',
-            }
-
-        self.stub_out('nova.volume.cinder.API.get', volume_get)
-        attempts = c._await_block_device_map_created(self.context, '1')
-        self.assertEqual(1, attempts)
-
     def test_await_block_device_created_retries_zero(self):
         c = self.compute
         self.flags(block_device_allocate_retries=0)
diff --git a/releasenotes/notes/block_device_allocate_retries-min-0-uprade-dc97b8f0e7716a3b.yaml b/releasenotes/notes/block_device_allocate_retries-min-0-uprade-dc97b8f0e7716a3b.yaml
new file mode 100644
index 000000000000..9cf02bb3cf69
--- /dev/null
+++ b/releasenotes/notes/block_device_allocate_retries-min-0-uprade-dc97b8f0e7716a3b.yaml
@@ -0,0 +1,6 @@
+---
+upgrade:
+  - |
+    The ``[DEFAULT]/block_device_allocate_retries`` configuration option now
+    has a minimum required value of 0. Any previous configuration with a value
+    less than zero will now result in an error.