Cache database and message queue connection objects

Recently in the gate we have seen a trace on some work-in-progress
patches:

  OperationalError: (pymysql.err.OperationalError)
                    (1040, u'Too many connections')

and at least one operator has reported that the number of database
connections increased significantly going from Mitaka to Newton.

It was suspected that the increase was caused by creating new oslo.db
transaction context managers on-the-fly when switching database
connections for cells. Comparing the dstat --tcp output of runs of the
gate-tempest-dsvm-neutron-full-ubuntu-xenial job with and without
caching of the database connections showed a difference of 445 active
TCP connections and 1495 active TCP connections, respectively [1].

This adds caching of the oslo.db transaction context managers and the
oslo.messaging transports to avoid creating a large number of objects
that are not being garbage-collected as expected.

Closes-Bug: #1691545

[1] https://docs.google.com/spreadsheets/d/1DIfFfX3kaA_SRoCM-aO7BN4IBEShChXLztOBFeKryt4/edit?usp=sharing

Change-Id: I17e0eb836dd87aac5859f506e7d771d42753d31a

1 files changed, 25 insertions, 5 deletions

diff --git a/nova/context.py b/nova/context.py
index be49d2945c..a22185dcbf 100644
--- a/nova/context.py
+++ b/nova/context.py
@@ -34,6 +34,10 @@ from nova import policy
 from nova import utils
 
 LOG = logging.getLogger(__name__)
+# TODO(melwitt): This cache should be cleared whenever WSGIService receives a
+# SIGHUP and periodically based on an expiration time. Currently, none of the
+# cell caches are purged, so neither is this one, for now.
+CELL_CACHE = {}
 
 
 class _ContextAuthPlugin(plugin.BaseAuthPlugin):
@@ -370,15 +374,31 @@ def set_target_cell(context, cell_mapping):
     :param context: The RequestContext to add connection information
     :param cell_mapping: An objects.CellMapping object or None
     """
+    global CELL_CACHE
     if cell_mapping is not None:
         # avoid circular import
         from nova import db
         from nova import rpc
-        db_connection_string = cell_mapping.database_connection
-        context.db_connection = db.create_context_manager(db_connection_string)
-        if not cell_mapping.transport_url.startswith('none'):
-            context.mq_connection = rpc.create_transport(
-                cell_mapping.transport_url)
+
+        # Synchronize access to the cache by multiple API workers.
+        @utils.synchronized(cell_mapping.uuid)
+        def get_or_set_cached_cell_and_set_connections():
+            try:
+                cell_tuple = CELL_CACHE[cell_mapping.uuid]
+            except KeyError:
+                db_connection_string = cell_mapping.database_connection
+                context.db_connection = db.create_context_manager(
+                    db_connection_string)
+                if not cell_mapping.transport_url.startswith('none'):
+                    context.mq_connection = rpc.create_transport(
+                        cell_mapping.transport_url)
+                CELL_CACHE[cell_mapping.uuid] = (context.db_connection,
+                                                 context.mq_connection)
+            else:
+                context.db_connection = cell_tuple[0]
+                context.mq_connection = cell_tuple[1]
+
+        get_or_set_cached_cell_and_set_connections()
     else:
         context.db_connection = None
         context.mq_connection = None