Connection to a ScyllaDB cluster is delayed as the driver tries to query system.peers_v2 table

The logic is now that if there is sharding information available, it's a Scylla cluster and then do NOT
try to use that table.

Fixes: scylladb#245

Signed-off-by: Yaniv Kaul <yaniv.kaul@scylladb.com>

Author: mykaul

cassandra/cluster.py

@@ -3668,6 +3668,11 @@ def _try_connect(self, host):
                   "registering watchers and refreshing schema and topology",
                   connection)
 
+        # Indirect way to determine if conencted to a ScyllaDB cluster, which does not support peers_v2
+        # If sharding information is available, it's a ScyllaDB cluster, so do not use peers_v2 table.
+        if connection.features.sharding_info is not None:
+            self._uses_peers_v2 = False
+
         # use weak references in both directions
         # _clear_watcher will be called when this ControlConnection is about to be finalized
         # _watch_callback will get the actual callback from the Connection and relay it to