Merge branch 'master' of github.com:mongodb/mongo-python-driver

This commit is contained in:
Steven Silvester 2025-02-06 13:37:40 -06:00
commit 2e6201d3c5
No known key found for this signature in database
GPG Key ID: B1BF5EC3A8B32F91
6 changed files with 627 additions and 34 deletions

View File

@ -35,10 +35,9 @@ jobs:
# https://github.com/github/feedback/discussions/7835#discussioncomment-1769026
buildplat:
- [ubuntu-20.04, "manylinux_x86_64", "cp3*-manylinux_x86_64"]
- [ubuntu-24.04-arm, "manylinux_aarch64", "cp3*-manylinux_aarch64"]
# Disabled pending PYTHON-5058
# - [ubuntu-24.04, "manylinux_ppc64le", "cp3*-manylinux_ppc64le"]
# - [ubuntu-24.04, "manylinux_s390x", "cp3*-manylinux_s390x"]
- [ubuntu-20.04, "manylinux_aarch64", "cp3*-manylinux_aarch64"]
- [ubuntu-20.04, "manylinux_ppc64le", "cp3*-manylinux_ppc64le"]
- [ubuntu-20.04, "manylinux_s390x", "cp3*-manylinux_s390x"]
- [ubuntu-20.04, "manylinux_i686", "cp3*-manylinux_i686"]
- [windows-2019, "win_amd6", "cp3*-win_amd64"]
- [windows-2019, "win32", "cp3*-win32"]
@ -63,6 +62,10 @@ jobs:
if: runner.os == 'Linux'
uses: docker/setup-qemu-action@v3
with:
# setup-qemu-action by default uses `tonistiigi/binfmt:latest` image,
# which is out of date. This causes seg faults during build.
# Here we manually fix the version.
image: tonistiigi/binfmt:qemu-v8.1.5
platforms: all
- name: Install cibuildwheel

View File

@ -0,0 +1,199 @@
# Copyright 2015-present MongoDB, Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Test AsyncMongoClient's mongos load balancing using a mock."""
from __future__ import annotations
import asyncio
import sys
import threading
from test.asynchronous.helpers import ConcurrentRunner
from pymongo.operations import _Op
sys.path[0:0] = [""]
from test.asynchronous import AsyncMockClientTest, async_client_context, connected, unittest
from test.asynchronous.pymongo_mocks import AsyncMockClient
from test.utils import async_wait_until
from pymongo.errors import AutoReconnect, InvalidOperation
from pymongo.server_selectors import writable_server_selector
from pymongo.topology_description import TOPOLOGY_TYPE
_IS_SYNC = False
class SimpleOp(ConcurrentRunner):
def __init__(self, client):
super().__init__()
self.client = client
self.passed = False
async def run(self):
await self.client.db.command("ping")
self.passed = True # No exception raised.
async def do_simple_op(client, ntasks):
tasks = [SimpleOp(client) for _ in range(ntasks)]
for t in tasks:
await t.start()
for t in tasks:
await t.join()
for t in tasks:
assert t.passed
async def writable_addresses(topology):
return {
server.description.address
for server in await topology.select_servers(writable_server_selector, _Op.TEST)
}
class TestMongosLoadBalancing(AsyncMockClientTest):
@async_client_context.require_connection
@async_client_context.require_no_load_balancer
async def asyncSetUp(self):
await super().asyncSetUp()
def mock_client(self, **kwargs):
mock_client = AsyncMockClient(
standalones=[],
members=[],
mongoses=["a:1", "b:2", "c:3"],
host="a:1,b:2,c:3",
connect=False,
**kwargs,
)
self.addAsyncCleanup(mock_client.aclose)
# Latencies in seconds.
mock_client.mock_rtts["a:1"] = 0.020
mock_client.mock_rtts["b:2"] = 0.025
mock_client.mock_rtts["c:3"] = 0.045
return mock_client
async def test_lazy_connect(self):
# While connected() ensures we can trigger connection from the main
# thread and wait for the monitors, this test triggers connection from
# several threads at once to check for data races.
nthreads = 10
client = self.mock_client()
self.assertEqual(0, len(client.nodes))
# Trigger initial connection.
await do_simple_op(client, nthreads)
await async_wait_until(lambda: len(client.nodes) == 3, "connect to all mongoses")
async def test_failover(self):
ntasks = 10
client = await connected(self.mock_client(localThresholdMS=0.001))
await async_wait_until(lambda: len(client.nodes) == 3, "connect to all mongoses")
# Our chosen mongos goes down.
client.kill_host("a:1")
# Trigger failover to higher-latency nodes. AutoReconnect should be
# raised at most once in each thread.
passed = []
async def f():
try:
await client.db.command("ping")
except AutoReconnect:
# Second attempt succeeds.
await client.db.command("ping")
passed.append(True)
tasks = [ConcurrentRunner(target=f) for _ in range(ntasks)]
for t in tasks:
await t.start()
for t in tasks:
await t.join()
self.assertEqual(ntasks, len(passed))
# Down host removed from list.
self.assertEqual(2, len(client.nodes))
async def test_local_threshold(self):
client = await connected(self.mock_client(localThresholdMS=30))
self.assertEqual(30, client.options.local_threshold_ms)
await async_wait_until(lambda: len(client.nodes) == 3, "connect to all mongoses")
topology = client._topology
# All are within a 30-ms latency window, see self.mock_client().
self.assertEqual({("a", 1), ("b", 2), ("c", 3)}, await writable_addresses(topology))
# No error
await client.admin.command("ping")
client = await connected(self.mock_client(localThresholdMS=0))
self.assertEqual(0, client.options.local_threshold_ms)
# No error
await client.db.command("ping")
# Our chosen mongos goes down.
client.kill_host("{}:{}".format(*next(iter(client.nodes))))
try:
await client.db.command("ping")
except:
pass
# We eventually connect to a new mongos.
async def connect_to_new_mongos():
try:
return await client.db.command("ping")
except AutoReconnect:
pass
await async_wait_until(connect_to_new_mongos, "connect to a new mongos")
async def test_load_balancing(self):
# Although the server selection JSON tests already prove that
# select_servers works for sharded topologies, here we do an end-to-end
# test of discovering servers' round trip times and configuring
# localThresholdMS.
client = await connected(self.mock_client())
await async_wait_until(lambda: len(client.nodes) == 3, "connect to all mongoses")
# Prohibited for topology type Sharded.
with self.assertRaises(InvalidOperation):
await client.address
topology = client._topology
self.assertEqual(TOPOLOGY_TYPE.Sharded, topology.description.topology_type)
# a and b are within the 15-ms latency window, see self.mock_client().
self.assertEqual({("a", 1), ("b", 2)}, await writable_addresses(topology))
client.mock_rtts["a:1"] = 0.045
# Discover only b is within latency window.
async def predicate():
return {("b", 2)} == await writable_addresses(topology)
await async_wait_until(
predicate,
'discover server "a" is too far',
)
if __name__ == "__main__":
unittest.main()

View File

@ -0,0 +1,374 @@
# Copyright 2016 MongoDB, Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Run the sdam monitoring spec tests."""
from __future__ import annotations
import asyncio
import json
import os
import sys
import time
from pathlib import Path
sys.path[0:0] = [""]
from test.asynchronous import AsyncIntegrationTest, async_client_context, client_knobs, unittest
from test.utils import (
ServerAndTopologyEventListener,
async_wait_until,
server_name_to_type,
)
from bson.json_util import object_hook
from pymongo import AsyncMongoClient, monitoring
from pymongo.asynchronous.collection import AsyncCollection
from pymongo.asynchronous.monitor import Monitor
from pymongo.common import clean_node
from pymongo.errors import ConnectionFailure, NotPrimaryError
from pymongo.hello import Hello
from pymongo.server_description import ServerDescription
from pymongo.topology_description import TOPOLOGY_TYPE
_IS_SYNC = False
# Location of JSON test specifications.
if _IS_SYNC:
TEST_PATH = os.path.join(Path(__file__).resolve().parent, "sdam_monitoring")
else:
TEST_PATH = os.path.join(Path(__file__).resolve().parent.parent, "sdam_monitoring")
def compare_server_descriptions(expected, actual):
if (expected["address"] != "{}:{}".format(*actual.address)) or (
server_name_to_type(expected["type"]) != actual.server_type
):
return False
expected_hosts = set(expected["arbiters"] + expected["passives"] + expected["hosts"])
return expected_hosts == {"{}:{}".format(*s) for s in actual.all_hosts}
def compare_topology_descriptions(expected, actual):
if TOPOLOGY_TYPE.__getattribute__(expected["topologyType"]) != actual.topology_type:
return False
expected = expected["servers"]
actual = actual.server_descriptions()
if len(expected) != len(actual):
return False
for exp_server in expected:
for _address, actual_server in actual.items():
if compare_server_descriptions(exp_server, actual_server):
break
else:
return False
return True
def compare_events(expected_dict, actual):
if not expected_dict:
return False, "Error: Bad expected value in YAML test"
if not actual:
return False, "Error: Event published was None"
expected_type, expected = list(expected_dict.items())[0]
if expected_type == "server_opening_event":
if not isinstance(actual, monitoring.ServerOpeningEvent):
return False, "Expected ServerOpeningEvent, got %s" % (actual.__class__)
if expected["address"] != "{}:{}".format(*actual.server_address):
return (
False,
"ServerOpeningEvent published with wrong address (expected" " {}, got {}".format(
expected["address"], actual.server_address
),
)
elif expected_type == "server_description_changed_event":
if not isinstance(actual, monitoring.ServerDescriptionChangedEvent):
return (False, "Expected ServerDescriptionChangedEvent, got %s" % (actual.__class__))
if expected["address"] != "{}:{}".format(*actual.server_address):
return (
False,
"ServerDescriptionChangedEvent has wrong address" " (expected {}, got {}".format(
expected["address"], actual.server_address
),
)
if not compare_server_descriptions(expected["newDescription"], actual.new_description):
return (False, "New ServerDescription incorrect in ServerDescriptionChangedEvent")
if not compare_server_descriptions(
expected["previousDescription"], actual.previous_description
):
return (
False,
"Previous ServerDescription incorrect in ServerDescriptionChangedEvent",
)
elif expected_type == "server_closed_event":
if not isinstance(actual, monitoring.ServerClosedEvent):
return False, "Expected ServerClosedEvent, got %s" % (actual.__class__)
if expected["address"] != "{}:{}".format(*actual.server_address):
return (
False,
"ServerClosedEvent published with wrong address" " (expected {}, got {}".format(
expected["address"], actual.server_address
),
)
elif expected_type == "topology_opening_event":
if not isinstance(actual, monitoring.TopologyOpenedEvent):
return False, "Expected TopologyOpenedEvent, got %s" % (actual.__class__)
elif expected_type == "topology_description_changed_event":
if not isinstance(actual, monitoring.TopologyDescriptionChangedEvent):
return (
False,
"Expected TopologyDescriptionChangedEvent, got %s" % (actual.__class__),
)
if not compare_topology_descriptions(expected["newDescription"], actual.new_description):
return (
False,
"New TopologyDescription incorrect in TopologyDescriptionChangedEvent",
)
if not compare_topology_descriptions(
expected["previousDescription"], actual.previous_description
):
return (
False,
"Previous TopologyDescription incorrect in TopologyDescriptionChangedEvent",
)
elif expected_type == "topology_await aclosed_event":
if not isinstance(actual, monitoring.TopologyClosedEvent):
return False, "Expected TopologyClosedEvent, got %s" % (actual.__class__)
else:
return False, f"Incorrect event: expected {expected_type}, actual {actual}"
return True, ""
def compare_multiple_events(i, expected_results, actual_results):
events_in_a_row = []
j = i
while j < len(expected_results) and isinstance(actual_results[j], actual_results[i].__class__):
events_in_a_row.append(actual_results[j])
j += 1
message = ""
for event in events_in_a_row:
for k in range(i, j):
passed, message = compare_events(expected_results[k], event)
if passed:
expected_results[k] = None
break
else:
return i, False, message
return j, True, ""
class TestAllScenarios(AsyncIntegrationTest):
async def asyncSetUp(self):
await super().asyncSetUp()
self.all_listener = ServerAndTopologyEventListener()
def create_test(scenario_def):
async def run_scenario(self):
with client_knobs(events_queue_frequency=0.05, min_heartbeat_interval=0.05):
await _run_scenario(self)
async def _run_scenario(self):
class NoopMonitor(Monitor):
"""Override the _run method to do nothing."""
async def _run(self):
await asyncio.sleep(0.05)
m = AsyncMongoClient(
host=scenario_def["uri"],
port=27017,
event_listeners=[self.all_listener],
_monitor_class=NoopMonitor,
)
topology = await m._get_topology()
try:
for phase in scenario_def["phases"]:
for source, response in phase.get("responses", []):
source_address = clean_node(source)
await topology.on_change(
ServerDescription(
address=source_address, hello=Hello(response), round_trip_time=0
)
)
expected_results = phase["outcome"]["events"]
expected_len = len(expected_results)
await async_wait_until(
lambda: len(self.all_listener.results) >= expected_len,
"publish all events",
timeout=15,
)
# Wait some time to catch possible lagging extra events.
await async_wait_until(lambda: topology._events.empty(), "publish lagging events")
i = 0
while i < expected_len:
result = (
self.all_listener.results[i] if len(self.all_listener.results) > i else None
)
# The order of ServerOpening/ClosedEvents doesn't matter
if isinstance(
result, (monitoring.ServerOpeningEvent, monitoring.ServerClosedEvent)
):
i, passed, message = compare_multiple_events(
i, expected_results, self.all_listener.results
)
self.assertTrue(passed, message)
else:
self.assertTrue(*compare_events(expected_results[i], result))
i += 1
# Assert no extra events.
extra_events = self.all_listener.results[expected_len:]
if extra_events:
self.fail(f"Extra events {extra_events!r}")
self.all_listener.reset()
finally:
await m.close()
return run_scenario
def create_tests():
for dirpath, _, filenames in os.walk(TEST_PATH):
for filename in filenames:
with open(os.path.join(dirpath, filename)) as scenario_stream:
scenario_def = json.load(scenario_stream, object_hook=object_hook)
# Construct test from scenario.
new_test = create_test(scenario_def)
test_name = f"test_{os.path.splitext(filename)[0]}"
new_test.__name__ = test_name
setattr(TestAllScenarios, new_test.__name__, new_test)
create_tests()
class TestSdamMonitoring(AsyncIntegrationTest):
knobs: client_knobs
listener: ServerAndTopologyEventListener
test_client: AsyncMongoClient
coll: AsyncCollection
@classmethod
def setUpClass(cls):
# Speed up the tests by decreasing the event publish frequency.
cls.knobs = client_knobs(
events_queue_frequency=0.1, heartbeat_frequency=0.1, min_heartbeat_interval=0.1
)
cls.knobs.enable()
cls.listener = ServerAndTopologyEventListener()
@classmethod
def tearDownClass(cls):
cls.knobs.disable()
@async_client_context.require_failCommand_fail_point
async def asyncSetUp(self):
await super().asyncSetUp()
retry_writes = async_client_context.supports_transactions()
self.test_client = await self.async_rs_or_single_client(
event_listeners=[self.listener], retryWrites=retry_writes
)
self.coll = self.test_client[self.client.db.name].test
await self.coll.insert_one({})
self.listener.reset()
async def asyncTearDown(self):
await super().asyncTearDown()
async def _test_app_error(self, fail_command_opts, expected_error):
address = await self.test_client.address
# Test that an application error causes a ServerDescriptionChangedEvent
# to be published.
data = {"failCommands": ["insert"]}
data.update(fail_command_opts)
fail_insert = {
"configureFailPoint": "failCommand",
"mode": {"times": 1},
"data": data,
}
async with self.fail_point(fail_insert):
if self.test_client.options.retry_writes:
await self.coll.insert_one({})
else:
with self.assertRaises(expected_error):
await self.coll.insert_one({})
await self.coll.insert_one({})
def marked_unknown(event):
return (
isinstance(event, monitoring.ServerDescriptionChangedEvent)
and event.server_address == address
and not event.new_description.is_server_type_known
)
def discovered_node(event):
return (
isinstance(event, monitoring.ServerDescriptionChangedEvent)
and event.server_address == address
and not event.previous_description.is_server_type_known
and event.new_description.is_server_type_known
)
def marked_unknown_and_rediscovered():
return (
len(self.listener.matching(marked_unknown)) >= 1
and len(self.listener.matching(discovered_node)) >= 1
)
# Topology events are not published synchronously
await async_wait_until(marked_unknown_and_rediscovered, "rediscover node")
# Expect a single ServerDescriptionChangedEvent for the network error.
marked_unknown_events = self.listener.matching(marked_unknown)
self.assertEqual(len(marked_unknown_events), 1, marked_unknown_events)
self.assertIsInstance(marked_unknown_events[0].new_description.error, expected_error)
async def test_network_error_publishes_events(self):
await self._test_app_error({"closeConnection": True}, ConnectionFailure)
# In 4.4+, not primary errors from failCommand don't cause SDAM state
# changes because topologyVersion is not incremented.
@async_client_context.require_version_max(4, 3)
async def test_not_primary_error_publishes_events(self):
await self._test_app_error(
{"errorCode": 10107, "closeConnection": False, "errorLabels": ["RetryableWriteError"]},
NotPrimaryError,
)
async def test_shutdown_error_publishes_events(self):
await self._test_app_error(
{"errorCode": 91, "closeConnection": False, "errorLabels": ["RetryableWriteError"]},
NotPrimaryError,
)
if __name__ == "__main__":
unittest.main()

View File

@ -15,8 +15,10 @@
"""Test MongoClient's mongos load balancing using a mock."""
from __future__ import annotations
import asyncio
import sys
import threading
from test.helpers import ConcurrentRunner
from pymongo.operations import _Op
@ -30,14 +32,10 @@ from pymongo.errors import AutoReconnect, InvalidOperation
from pymongo.server_selectors import writable_server_selector
from pymongo.topology_description import TOPOLOGY_TYPE
@client_context.require_connection
@client_context.require_no_load_balancer
def setUpModule():
pass
_IS_SYNC = True
class SimpleOp(threading.Thread):
class SimpleOp(ConcurrentRunner):
def __init__(self, client):
super().__init__()
self.client = client
@ -48,15 +46,15 @@ class SimpleOp(threading.Thread):
self.passed = True # No exception raised.
def do_simple_op(client, nthreads):
threads = [SimpleOp(client) for _ in range(nthreads)]
for t in threads:
def do_simple_op(client, ntasks):
tasks = [SimpleOp(client) for _ in range(ntasks)]
for t in tasks:
t.start()
for t in threads:
for t in tasks:
t.join()
for t in threads:
for t in tasks:
assert t.passed
@ -68,6 +66,11 @@ def writable_addresses(topology):
class TestMongosLoadBalancing(MockClientTest):
@client_context.require_connection
@client_context.require_no_load_balancer
def setUp(self):
super().setUp()
def mock_client(self, **kwargs):
mock_client = MockClient(
standalones=[],
@ -98,7 +101,7 @@ class TestMongosLoadBalancing(MockClientTest):
wait_until(lambda: len(client.nodes) == 3, "connect to all mongoses")
def test_failover(self):
nthreads = 10
ntasks = 10
client = connected(self.mock_client(localThresholdMS=0.001))
wait_until(lambda: len(client.nodes) == 3, "connect to all mongoses")
@ -118,14 +121,14 @@ class TestMongosLoadBalancing(MockClientTest):
passed.append(True)
threads = [threading.Thread(target=f) for _ in range(nthreads)]
for t in threads:
tasks = [ConcurrentRunner(target=f) for _ in range(ntasks)]
for t in tasks:
t.start()
for t in threads:
for t in tasks:
t.join()
self.assertEqual(nthreads, len(passed))
self.assertEqual(ntasks, len(passed))
# Down host removed from list.
self.assertEqual(2, len(client.nodes))
@ -183,8 +186,11 @@ class TestMongosLoadBalancing(MockClientTest):
client.mock_rtts["a:1"] = 0.045
# Discover only b is within latency window.
def predicate():
return {("b", 2)} == writable_addresses(topology)
wait_until(
lambda: {("b", 2)} == writable_addresses(topology),
predicate,
'discover server "a" is too far',
)

View File

@ -15,10 +15,12 @@
"""Run the sdam monitoring spec tests."""
from __future__ import annotations
import asyncio
import json
import os
import sys
import time
from pathlib import Path
sys.path[0:0] = [""]
@ -39,8 +41,13 @@ from pymongo.synchronous.collection import Collection
from pymongo.synchronous.monitor import Monitor
from pymongo.topology_description import TOPOLOGY_TYPE
_IS_SYNC = True
# Location of JSON test specifications.
_TEST_PATH = os.path.join(os.path.dirname(os.path.realpath(__file__)), "sdam_monitoring")
if _IS_SYNC:
TEST_PATH = os.path.join(Path(__file__).resolve().parent, "sdam_monitoring")
else:
TEST_PATH = os.path.join(Path(__file__).resolve().parent.parent, "sdam_monitoring")
def compare_server_descriptions(expected, actual):
@ -247,7 +254,7 @@ def create_test(scenario_def):
def create_tests():
for dirpath, _, filenames in os.walk(_TEST_PATH):
for dirpath, _, filenames in os.walk(TEST_PATH):
for filename in filenames:
with open(os.path.join(dirpath, filename)) as scenario_stream:
scenario_def = json.load(scenario_stream, object_hook=object_hook)
@ -268,31 +275,33 @@ class TestSdamMonitoring(IntegrationTest):
coll: Collection
@classmethod
@client_context.require_failCommand_fail_point
def setUpClass(cls):
super().setUp(cls)
# Speed up the tests by decreasing the event publish frequency.
cls.knobs = client_knobs(
events_queue_frequency=0.1, heartbeat_frequency=0.1, min_heartbeat_interval=0.1
)
cls.knobs.enable()
cls.listener = ServerAndTopologyEventListener()
retry_writes = client_context.supports_transactions()
cls.test_client = cls.unmanaged_rs_or_single_client(
event_listeners=[cls.listener], retryWrites=retry_writes
)
cls.coll = cls.test_client[cls.client.db.name].test
cls.coll.insert_one({})
@classmethod
def tearDownClass(cls):
cls.test_client.close()
cls.knobs.disable()
super().tearDownClass()
@client_context.require_failCommand_fail_point
def setUp(self):
super().setUp()
retry_writes = client_context.supports_transactions()
self.test_client = self.rs_or_single_client(
event_listeners=[self.listener], retryWrites=retry_writes
)
self.coll = self.test_client[self.client.db.name].test
self.coll.insert_one({})
self.listener.reset()
def tearDown(self):
super().tearDown()
def _test_app_error(self, fail_command_opts, expected_error):
address = self.test_client.address
@ -334,7 +343,7 @@ class TestSdamMonitoring(IntegrationTest):
and len(self.listener.matching(discovered_node)) >= 1
)
# Topology events are published asynchronously
# Topology events are not published synchronously
wait_until(marked_unknown_and_rediscovered, "rediscover node")
# Expect a single ServerDescriptionChangedEvent for the network error.

View File

@ -221,6 +221,7 @@ converted_tests = [
"test_logger.py",
"test_max_staleness.py",
"test_monitoring.py",
"test_mongos_load_balancing.py",
"test_on_demand_csfle.py",
"test_raw_bson.py",
"test_read_concern.py",
@ -231,6 +232,7 @@ converted_tests = [
"test_retryable_writes.py",
"test_retryable_writes_unified.py",
"test_run_command.py",
"test_sdam_monitoring_spec.py",
"test_server_selection_logging.py",
"test_session.py",
"test_server_selection_rtt.py",