openstack/neutron
Code Issues Proposed changes

[FT] Initialize the segment ranges sequentially

Browse Source 
In the functional tests
``test_initialize_nsrange_support_parallel_exec_*_init_time``, the
segment ranges initialization is done sequentially to avoid a race
condition with the database driver.

In a real environment, only the first WSGI worker thread is in charge
of initializing these registers; in a multi server environment, the
possibility of race condition between then is almost none and should
be solved by the database engine.

Closes-Bug: #2130564

Signed-off-by: Rodolfo Alonso Hernandez <ralonsoh@redhat.com>
Change-Id: Iedf74efb3b970cca8c9bee749d9136d8dcd140b3
This commit is contained in:
Rodolfo Alonso Hernandez
2025-11-03 14:06:33 +00:00
parent 10c6a610ea
commit 3137a97f46
1 changed files with 8 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
neutron/tests/functional/plugins/ml2/drivers/test_type_tunnel.py
View File

@@ -14,6 +14,7 @@
# under the License.
from concurrent import futures
import time
from neutron_lib import constants
from neutron_lib import context
@@ -29,14 +30,18 @@ from neutron.plugins.ml2.drivers import type_geneve
from neutron.tests.unit import testlib_api
def _initialize_network_segment_range_support(type_driver, start_time):
def _initialize_network_segment_range_support(type_driver, worker_num,
same_init_time):
# This method is similar to
# ``_TunnelTypeDriverBase.initialize_network_segment_range_support``.
# The method first deletes the existing default network ranges and then
# creates the new ones. It also adds an extra second before closing the
# DB transaction.
#
start_time = worker_num if not same_init_time else 0
admin_context = context.get_admin_context()
try:
time.sleep(worker_num / 4)
with db_api.CONTEXT_WRITER.using(admin_context):
type_driver._delete_expired_default_network_segment_ranges(
admin_context, start_time)
@@ -84,17 +89,10 @@ class TunnelTypeDriverBaseTestCase(testlib_api.MySQLTestCaseMixin,
max_workers = 3
_futures = []
with futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
if same_init_time:
# All workers are started at the same init time.
for idx in range(max_workers):
_futures.append(executor.submit(
_initialize_network_segment_range_support,
self.type_driver, 0))
else:
# All workers have different init times.
for idx in range(max_workers):
_futures.append(executor.submit(
_initialize_network_segment_range_support,
self.type_driver, idx))
self.type_driver, idx, same_init_time))
for _future in _futures:
_future.result()