Fix handling of circular MOVED redirects in cluster slot mapping

redis/asyncio/cluster.py

@@ -1456,27 +1456,30 @@ def _update_moved_slots(self) -> None:
                 e.host, e.port, PRIMARY, **self.connection_kwargs
             )
             self.set_nodes(self.nodes_cache, {redirected_node.name: redirected_node})
-        if redirected_node in self.slots_cache[e.slot_id]:
+        slot_nodes = self.slots_cache[e.slot_id]
+        if redirected_node not in slot_nodes:
+            # The new slot owner is a new server, or a server from a different
+            # shard. We need to remove all current nodes from the slot's list
+            # (including replications) and add just the new node.
+            self.slots_cache[e.slot_id] = [redirected_node]
+        elif redirected_node is not slot_nodes[0]:
             # The MOVED error resulted from a failover, and the new slot owner
             # had previously been a replica.
-            old_primary = self.slots_cache[e.slot_id][0]
+            old_primary = slot_nodes[0]
             # Update the old primary to be a replica and add it to the end of
             # the slot's node list
             old_primary.server_type = REPLICA
-            self.slots_cache[e.slot_id].append(old_primary)
+            slot_nodes.append(old_primary)
             # Remove the old replica, which is now a primary, from the slot's
             # node list
-            self.slots_cache[e.slot_id].remove(redirected_node)
+            slot_nodes.remove(redirected_node)
             # Override the old primary with the new one
-            self.slots_cache[e.slot_id][0] = redirected_node
+            slot_nodes[0] = redirected_node
             if self.default_node == old_primary:
                 # Update the default node with the new primary
                 self.default_node = redirected_node
-        else:
-            # The new slot owner is a new server, or a server from a different
-            # shard. We need to remove all current nodes from the slot's list
-            # (including replications) and add just the new node.
-            self.slots_cache[e.slot_id] = [redirected_node]
+        # else: circular MOVED to current primary -> no-op
+
         # Reset moved_exception
         self._moved_exception = None
 

redis/cluster.py

@@ -1693,27 +1693,30 @@ def _update_moved_slots(self):
             # This is a new node, we will add it to the nodes cache
             redirected_node = ClusterNode(e.host, e.port, PRIMARY)
             self.nodes_cache[redirected_node.name] = redirected_node
-        if redirected_node in self.slots_cache[e.slot_id]:
+        slot_nodes = self.slots_cache[e.slot_id]
+        if redirected_node not in slot_nodes:
+            # The new slot owner is a new server, or a server from a different
+            # shard. We need to remove all current nodes from the slot's list
+            # (including replications) and add just the new node.
+            self.slots_cache[e.slot_id] = [redirected_node]
+        elif redirected_node is not slot_nodes[0]:
             # The MOVED error resulted from a failover, and the new slot owner
             # had previously been a replica.
-            old_primary = self.slots_cache[e.slot_id][0]
+            old_primary = slot_nodes[0]
             # Update the old primary to be a replica and add it to the end of
             # the slot's node list
             old_primary.server_type = REPLICA
-            self.slots_cache[e.slot_id].append(old_primary)
+            slot_nodes.append(old_primary)
             # Remove the old replica, which is now a primary, from the slot's
             # node list
-            self.slots_cache[e.slot_id].remove(redirected_node)
+            slot_nodes.remove(redirected_node)
             # Override the old primary with the new one
-            self.slots_cache[e.slot_id][0] = redirected_node
+            slot_nodes[0] = redirected_node
             if self.default_node == old_primary:
                 # Update the default node with the new primary
                 self.default_node = redirected_node
-        else:
-            # The new slot owner is a new server, or a server from a different
-            # shard. We need to remove all current nodes from the slot's list
-            # (including replications) and add just the new node.
-            self.slots_cache[e.slot_id] = [redirected_node]
+        # else: circular MOVED to current primary -> no-op
+
         # Reset moved_exception
         self._moved_exception = None
 

tests/test_asyncio/test_cluster.py

@@ -233,19 +233,25 @@ def mock_all_nodes_resp(rc: RedisCluster, response: Any) -> RedisCluster:
 
 
 async def moved_redirection_helper(
-    create_redis: Callable[..., RedisCluster], failover: bool = False
+    create_redis: Callable[..., RedisCluster],
+    failover: bool = False,
+    circular_moved=False,
 ) -> None:
     """
-    Test that the client handles MOVED response after a failover.
-    Redirection after a failover means that the redirection address is of a
-    replica that was promoted to a primary.
+    Test that the client correctly handles MOVED responses in the following scenarios:
+    1.	Slot migration to a different shard (failover=False, circular_moved=False) —
+        a standard slot move between shards.
+    2.	Failover event (failover=True, circular_moved=False) —
+        the redirect target is a replica that has just been promoted to primary.
+    3.	Circular MOVED (failover=False, circular_moved=True) —
+        the redirect points to a node already known to be the primary of its shard.
 
     At first call it should return a MOVED ResponseError that will point
     the client to the next server it should talk to.
 
     Verify that:
     1. it tries to talk to the redirected node
-    2. it updates the slot's primary to the redirected node
+    2. it updates the slot's primary to the redirected node, if required
 
     For a failover, also verify:
     3. the redirected node's server type updated to 'primary'
@@ -261,6 +267,8 @@ async def moved_redirection_helper(
             warnings.warn("Skipping this test since it requires to have a replica")
             return
         redirect_node = rc.nodes_manager.slots_cache[slot][1]
+    elif circular_moved:
+        redirect_node = prev_primary
     else:
         # Use one of the primaries to be the redirected node
         redirect_node = rc.get_primaries()[0]
@@ -287,6 +295,10 @@ def ok_response(self, *args, **options):
         if failover:
             assert rc.get_node(host=r_host, port=r_port).server_type == PRIMARY
             assert prev_primary.server_type == REPLICA
+        elif circular_moved:
+            fetched_node = rc.get_node(host=r_host, port=r_port)
+            assert fetched_node == prev_primary
+            assert fetched_node.server_type == PRIMARY
 
 
 class TestRedisClusterObj:
@@ -613,6 +625,17 @@ async def test_moved_redirection_after_failover(
         """
         await moved_redirection_helper(create_redis, failover=True)
 
+    async def test_moved_redirection_circular_moved(
+        self, create_redis: Callable[..., RedisCluster]
+    ) -> None:
+        """
+        Verify that the client does not update its slot map when receiving a circular MOVED response
+        (i.e., a MOVED redirect pointing back to the same node), and retries again the same node.
+        """
+        await moved_redirection_helper(
+            create_redis, failover=False, circular_moved=True
+        )
+
     async def test_refresh_using_specific_nodes(
         self, create_redis: Callable[..., RedisCluster]
     ) -> None:

tests/test_cluster.py

@@ -273,18 +273,22 @@ def find_node_ip_based_on_port(cluster_client, port):
             return node.host
 
 
-def moved_redirection_helper(request, failover=False):
+def moved_redirection_helper(request, failover=False, circular_moved=False):
     """
-    Test that the client handles MOVED response after a failover.
-    Redirection after a failover means that the redirection address is of a
-    replica that was promoted to a primary.
+    Test that the client correctly handles MOVED responses in the following scenarios:
+    1.	Slot migration to a different shard (failover=False, circular_moved=False) —
+        a standard slot move between shards.
+    2.	Failover event (failover=True, circular_moved=False) —
+        the redirect target is a replica that has just been promoted to primary.
+    3.	Circular MOVED (failover=False, circular_moved=True) —
+        the redirect points to a node already known to be the primary of its shard.
 
     At first call it should return a MOVED ResponseError that will point
     the client to the next server it should talk to.
 
     Verify that:
     1. it tries to talk to the redirected node
-    2. it updates the slot's primary to the redirected node
+    2. it updates the slot's primary to the redirected node, if required
 
     For a failover, also verify:
     3. the redirected node's server type updated to 'primary'
@@ -300,8 +304,10 @@ def moved_redirection_helper(request, failover=False):
             warnings.warn("Skipping this test since it requires to have a replica")
             return
         redirect_node = rc.nodes_manager.slots_cache[slot][1]
+    elif circular_moved:
+        redirect_node = prev_primary
     else:
-        # Use one of the primaries to be the redirected node
+        # Use one of the other primaries to be the redirected node
         redirect_node = rc.get_primaries()[0]
     r_host = redirect_node.host
     r_port = redirect_node.port
@@ -324,6 +330,10 @@ def ok_response(connection, *args, **options):
         if failover:
             assert rc.get_node(host=r_host, port=r_port).server_type == PRIMARY
             assert prev_primary.server_type == REPLICA
+        elif circular_moved:
+            fetched_node = rc.get_node(host=r_host, port=r_port)
+            assert fetched_node == prev_primary
+            assert fetched_node.server_type == PRIMARY
 
 
 @pytest.mark.onlycluster
@@ -547,6 +557,13 @@ def test_moved_redirection_after_failover(self, request):
         """
         moved_redirection_helper(request, failover=True)
 
+    def test_moved_redirection_circular_moved(self, request):
+        """
+        Verify that the client does not update its slot map when receiving a circular MOVED response
+        (i.e., a MOVED redirect pointing back to the same node), and retries again the same node.
+        """
+        moved_redirection_helper(request, failover=False, circular_moved=True)
+
     def test_refresh_using_specific_nodes(self, request):
         """
         Test making calls on specific nodes when the cluster has failed over to