diff --git a/etc/nsx.ini b/etc/nsx.ini
index 5159d271bb..053720cadf 100644
--- a/etc/nsx.ini
+++ b/etc/nsx.ini
@@ -369,7 +369,7 @@
 
 # The amount of time in seconds to wait before ensuring connectivity to
 # the NSX manager if no manager connection has been used.
-# conn_idle_timeout = 60
+# conn_idle_timeout = 10
 
 # UUID of the default tier0 router that will be used for connecting to
 # tier1 logical routers and configuring external networks
diff --git a/vmware_nsx/common/config.py b/vmware_nsx/common/config.py
index 6dd7989a21..9fd1772f28 100644
--- a/vmware_nsx/common/config.py
+++ b/vmware_nsx/common/config.py
@@ -232,7 +232,7 @@ nsx_v3_opts = [
                help=_("Maximum concurrent connections to each NSX "
                       "manager.")),
     cfg.IntOpt('conn_idle_timeout',
-               default=60,
+               default=10,
                help=_('Ensure connectivity to the NSX manager if a connection '
                       'is not used within timeout seconds.')),
     cfg.IntOpt('redirects',
diff --git a/vmware_nsx/nsxlib/v3/cluster.py b/vmware_nsx/nsxlib/v3/cluster.py
index ec52c3c6e2..988861b655 100644
--- a/vmware_nsx/nsxlib/v3/cluster.py
+++ b/vmware_nsx/nsxlib/v3/cluster.py
@@ -326,11 +326,6 @@ class ClusteredAPI(object):
                 if up == len(self._endpoints)
                 else ClusterHealth.ORANGE)
 
-    def revalidate_endpoints(self):
-        # validate each endpoint in serial
-        for endpoint in self._endpoints.values():
-            self._validate(endpoint)
-
     def _validate(self, endpoint):
         try:
             with endpoint.pool.item() as conn:
@@ -343,7 +338,7 @@ class ClusteredAPI(object):
                             "'%(ep)s' due to: %(err)s"),
                         {'ep': endpoint, 'err': e})
 
-    def _select_endpoint(self, revalidate=False):
+    def _select_endpoint(self):
         connected = {}
         for provider_id, endpoint in self._endpoints.items():
             if endpoint.state == EndpointState.UP:
@@ -352,12 +347,6 @@ class ClusteredAPI(object):
                     # connection can be used now
                     return endpoint
 
-        if not connected and revalidate:
-            LOG.debug("All endpoints DOWN; revalidating.")
-            # endpoints may have become available, try to revalidate
-            self.revalidate_endpoints()
-            return self._select_endpoint(revalidate=False)
-
         # no free connections; randomly select a connected endpoint
         # which will likely wait on pool.item() until a connection frees up
         return (connected[random.choice(connected.keys())]
@@ -382,8 +371,10 @@ class ClusteredAPI(object):
 
     @contextlib.contextmanager
     def endpoint_connection(self):
-        endpoint = self._select_endpoint(revalidate=True)
+        endpoint = self._select_endpoint()
         if not endpoint:
+            # all endpoints are DOWN and will have their next
+            # state updated as per _endpoint_keepalive()
             raise nsx_exc.ServiceClusterUnavailable(
                 cluster_id=self.cluster_id)
 
diff --git a/vmware_nsx/tests/unit/nsx_v3/test_plugin.py b/vmware_nsx/tests/unit/nsx_v3/test_plugin.py
index 376b5de2f3..d42434d27e 100644
--- a/vmware_nsx/tests/unit/nsx_v3/test_plugin.py
+++ b/vmware_nsx/tests/unit/nsx_v3/test_plugin.py
@@ -71,8 +71,6 @@ class NsxV3PluginTestCaseMixin(test_plugin.NeutronDbPluginV2TestCase,
         self.cluster = nsx_cluster.NSXClusteredAPI(
             http_provider=nsxlib_testcase.MemoryMockAPIProvider(self.mock_api))
 
-        self.cluster.revalidate_endpoints()
-
         def _patch_object(*args, **kwargs):
             patcher = mock.patch.object(*args, **kwargs)
             patcher.start()