Change prepare_on_all_hosts default to False

CHANGELOG.rst

@@ -27,6 +27,10 @@ Others
   come through unchanged. Previously ``DRIVER_NAME`` and ``DRIVER_VERSION`` could be
   overridden, which misreported the driver to the server for the life of the connection
   and, in the clients table, to the operator reading the row.
+* ``Cluster.prepare_on_all_hosts`` now defaults to ``False``. In multi-DC deployments eager
+  preparation previously ran on every pooled host, including remote hosts that are rarely or
+  never queried. Disabling it by default avoids that waste; an ``UNPREPARED`` response still
+  triggers on-demand reprepare and retry, so correctness is unaffected.
 * ``PreparedStatement.result_metadata`` and ``PreparedStatement.result_metadata_id`` are
   now read-only. They are replaced together by
   ``PreparedStatement.update_result_metadata()``, so a request can never observe a metadata

cassandra/cluster.py

@@ -984,11 +984,14 @@ def default_retry_policy(self, policy):
     establish connection pools. This can cause a rush of connections and queries if not mitigated with this factor.
     """
 
-    prepare_on_all_hosts = True
+    prepare_on_all_hosts = False
     """
     Specifies whether statements should be prepared on all hosts, or just one.
 
-    This can reasonably be disabled on long-running applications with numerous clients preparing statements on startup,
+    When enabled, statements are eagerly prepared on every host with an open connection pool. In multi-DC
+    deployments this includes remote hosts that are rarely or never queried on the happy path; preparing on them
+    is purely a latency optimization, since an ``UNPREPARED`` response always triggers on-demand reprepare and
+    retry. It can be enabled on long-running applications with numerous clients preparing statements on startup,
     where a randomized initial condition of the load balancing policy can be expected to distribute prepares from
     different clients across the cluster.
     """
@@ -1204,7 +1207,7 @@ def __init__(self,
                  schema_metadata_page_size=1000,
                  address_translator=None,
                  status_event_refresh_window=2,
-                 prepare_on_all_hosts=True,
+                 prepare_on_all_hosts=False,
                  reprepare_on_up=True,
                  execution_profiles=None,
                  allow_beta_protocol_version=False,

tests/integration/standard/test_shard_aware.py

@@ -71,7 +71,10 @@ def verify_same_shard_in_tracing(self, results, shard_name):
             assert shard_name in event.thread_name
         assert 'querying locally' in "\n".join([event.description for event in events])
 
-        trace_id = results.response_future.get_query_trace_ids()[0]
+        # Use the last trace id: prepare_on_all_hosts defaults to False now, so a query
+        # against a host that hasn't prepared the statement yet can get UNPREPARED and
+        # retry, which appends an earlier, incomplete trace before the one that matters.
+        trace_id = results.response_future.get_query_trace_ids()[-1]
         traces = self.session.execute("SELECT * FROM system_traces.events WHERE session_id = %s", (trace_id,))
         events = [event for event in traces]
         for event in events:

tests/integration/standard/test_tablets.py

@@ -37,7 +37,10 @@ def verify_hosts_in_tracing(self, results, expected):
         assert len(host_set) == expected
         assert 'locally' in "\n".join([event.description for event in events])
 
-        trace_id = results.response_future.get_query_trace_ids()[0]
+        # Use the last trace id: prepare_on_all_hosts defaults to False now, so a query
+        # against a host that hasn't prepared the statement yet can get UNPREPARED and
+        # retry, which appends an earlier, incomplete trace before the one that matters.
+        trace_id = results.response_future.get_query_trace_ids()[-1]
         traces = self.session.execute("SELECT * FROM system_traces.events WHERE session_id = %s", (trace_id,))
         events = [event for event in traces]
         host_set = set()
@@ -63,7 +66,8 @@ def verify_same_shard_in_tracing(self, results):
         assert len(shard_set) == 1
         assert 'locally' in "\n".join([event.description for event in events])
 
-        trace_id = results.response_future.get_query_trace_ids()[0]
+        # See verify_hosts_in_tracing: use the last trace id, not the first.
+        trace_id = results.response_future.get_query_trace_ids()[-1]
         traces = self.session.execute("SELECT * FROM system_traces.events WHERE session_id = %s", (trace_id,))
         events = [event for event in traces]
         shard_set = set()