litellm/tests/integration/conftest.py
devin-ai-integration[bot] 02feb67c20
test(integration): cut CircleCI proxy boot time with concurrent worker stacks and single-worker test proxies (#46037)
* test(integration): build worker stacks concurrently and record proxy boot phases

The controller built every xdist worker's database, Redis, upstream and gateway
proxy one after another before any test ran, about 95 s per extensions shard,
and since every multi-worker suite is now isolated the same serial wait hits
mcp, security and cost too. The stacks now start on a thread pool bounded by
the CPU count. Each stack reads its endpoints and results directory from an
explicit environment instead of a patched os.environ, every stack that started
is registered for cleanup, and a failure in one stops the others before the
error is raised

Owned proxy startup.json now records the wall-clock spawn time, the worker
count and when the Prisma CLI child ran, when the port started accepting and
when the first worker answered readiness, and exit.json records when the stop
started and ended, so the boot time can be split and overlapping gateway boots
can be seen in the artifacts

* test(integration): boot one uvicorn worker where a test does not assert cross-worker behaviour

A 2-worker owned proxy takes about twice as long to become ready as a 1-worker
one, because the uvicorn parent imports the app and each spawned worker imports
it again. The passthrough upstream error visibility cases, the single-request
S3 dedupe cases and the single-request spend log payload cases send one or two
requests and assert on the response, the proxy log and the spend row, none of
which depends on a second worker, so they now boot the default single worker

Cases that do rely on two workers keep them, the concurrent S3 burst through a
sink outage, previous_response_id replay, response cache hits, the concurrent
spend bursts and upstream stop and recovery

* test(observability): settle OTel excluded-services traces by a quiet period instead of a full window

_trace_spans slept through its whole 15 or 30 s window on every call even when
the trace had long stopped growing. It now polls and returns once the trace
has spans and has not gained one for five tenant export intervals, and the
module pins the tenant batch exporter's schedule delay to 1 s so that bound is
set by the test rather than by the SDK default. Absence checks still run only
after the trace stopped growing, and against a build that leaks postgres spans
to the tenant they still fail

These cases also boot one uvicorn worker now. None of them asserts anything
about a second worker

---------

Co-authored-by: Yuneng Jiang <yuneng@berri.ai>
2026-10-12 09:48:42 -07:00

376 lines
16 KiB
Python

from __future__ import annotations
import hashlib
import json
import os
import re
from collections.abc import AsyncIterator, Generator, Iterator, Mapping, Sequence
from contextlib import ExitStack
from copy import copy
from functools import partial
from importlib.metadata import version
from pathlib import Path
from typing import TYPE_CHECKING, Final, Protocol, runtime_checkable
import httpx
import pytest
import pytest_asyncio
from execnet import XSpec
from pydantic import TypeAdapter
from redis import Redis
from tests.integration._support.client import Gateway, Scenario, eventually, gateway_from_environment
from tests.integration._support.generation import LIFECYCLE_SETTINGS
from tests.integration._support.manifest import OWNED_DIRECTORIES
from tests.integration._support.ordering import collected_case, fixture_groups, scheduling_groups
from tests.integration._support.provider import SharedProvider, shared_provider
from tests.integration._support.routing import RoutingPlugin
from tests.integration._support.workers import gateway_profile, start_worker_stacks, worker_stack
from tests.integration.run import GITHUB_FILES
if TYPE_CHECKING:
from tests.integration._support.actors import ActorSet, GatewayClient, ManagementDatabase, ScratchResources
COLLECTED: Final = pytest.StashKey[tuple[str, ...]]()
INVENTORY: Final = pytest.StashKey[tuple[str, ...]]()
REPORTS: Final = pytest.StashKey[list[pytest.TestReport]]()
WORKER_INVENTORY: Final = TypeAdapter(tuple[str, ...])
SHARD_PATH: Final[TypeAdapter[Path | None]] = TypeAdapter(Path | None)
SHARD_FILES: Final = pytest.StashKey[frozenset[str] | None]()
GROUP_SUFFIX: Final = re.compile(r"@integration_fixture_[0-9a-f]{64}$")
INTEGER_OPTION: Final = TypeAdapter(int)
STRING_ENVIRONMENT: Final = TypeAdapter(dict[str, str])
@runtime_checkable
class WorkerConfig(Protocol):
@property
def workeroutput(self) -> dict[str, object]: ...
class WorkerNode(Protocol):
@property
def workeroutput(self) -> Mapping[str, object]: ...
def _worker_output(config: object) -> dict[str, object] | None:
return config.workeroutput if isinstance(config, WorkerConfig) else None
def pytest_addoption(parser: pytest.Parser) -> None:
parser.addoption("--integration-order-seed", type=int, default=0)
parser.addoption("--integration-shard-count", type=int, default=1)
parser.addoption("--integration-shard-index", type=int, default=0)
parser.addoption("--integration-shard-files", type=Path)
def pytest_configure(config: pytest.Config) -> None:
config.addinivalue_line("markers", "integration: owned real-service integration contracts")
config.addinivalue_line("markers", "covers(*ids): legacy contract IDs kept for existing tests, not enforced")
config.stash[REPORTS] = []
config.stash[SHARD_FILES] = _shard_files(SHARD_PATH.validate_python(config.getoption("integration_shard_files")))
config.pluginmanager.register(IntegrationReportPlugin(config))
if os.environ.get("INTEGRATION_ROUTING"):
config.pluginmanager.register(RoutingPlugin(config))
@pytest.hookimpl(optionalhook=True)
def pytest_xdist_setupnodes(config: pytest.Config, specs: Sequence[XSpec]) -> None:
if (
os.environ.get("INTEGRATION_WORKER_ISOLATION") != "1"
or config.getoption("collectonly") is True
or config.stash[SHARD_FILES] == frozenset()
):
return
resources: Final = ExitStack()
config.add_cleanup(resources.close)
destination: Final = Path(os.environ["INTEGRATION_RESULTS_DIR"])
stacks: Final = start_worker_stacks(
tuple(destination / f"gw{index}" for index in range(len(specs))),
partial(worker_stack, profile=gateway_profile(os.environ)),
resources,
os.cpu_count() or 1,
)
for spec, stack in zip(specs, stacks, strict=True):
spec.env = {**STRING_ENVIRONMENT.validate_python(spec.env), **stack}
class IntegrationReportPlugin:
def __init__(self, config: pytest.Config) -> None:
self.config = config
def pytest_runtest_logreport(self, report: pytest.TestReport) -> None:
self.config.stash[REPORTS].append(report)
@pytest.hookimpl(optionalhook=True)
def pytest_xdist_node_collection_finished(self, node: object, ids: Sequence[str]) -> None:
self.config.stash[COLLECTED] = tuple(_case_identity(nodeid) for nodeid in ids if _owned(nodeid))
@pytest.hookimpl(optionalhook=True)
def pytest_testnodedown(self, node: WorkerNode, error: object) -> None:
if error is not None:
return
inventory: Final = WORKER_INVENTORY.validate_python(node.workeroutput.get("integration_inventory"))
previous: Final = self.config.stash.get(INVENTORY, inventory)
if inventory != previous:
raise pytest.UsageError("Integration worker collection inventories disagree")
self.config.stash[INVENTORY] = inventory
def pytest_sessionfinish(self, session: pytest.Session, exitstatus: int) -> None:
output: Final = _worker_output(self.config)
if output is not None:
output["integration_inventory"] = self.config.stash.get(INVENTORY, ())
def _case_identity(nodeid: str) -> str:
return GROUP_SUFFIX.sub("", nodeid)
@pytest.hookimpl(wrapper=True, optionalhook=True)
def pytest_report_from_serializable() -> Generator[
None, pytest.TestReport | pytest.CollectReport | None, pytest.TestReport | pytest.CollectReport | None
]:
report: Final = yield
if not isinstance(report, pytest.TestReport):
return report
normalized: Final = copy(report)
normalized.nodeid = _case_identity(report.nodeid)
return normalized
@pytest.hookimpl(wrapper=True, optionalhook=True)
def pytest_handlecrashitem(report: pytest.TestReport) -> Generator[None, None, None]:
result: Final = yield
report.nodeid = _case_identity(report.nodeid) # rebind-ok: xdist forwards this same report after the crash hook
return result
def _owned(nodeid: str) -> bool:
parts: Final = Path(nodeid.split("::", 1)[0]).parts
return parts[:2] == ("tests", "integration") and len(parts) > 3 and parts[2] in OWNED_DIRECTORIES
def _digest(seed: int, identity: str) -> bytes:
return hashlib.sha256(f"{seed}:{identity}".encode()).digest()
def _order_key(seed: int, nodeid: str, group: str = "", inner_group: str = "") -> tuple[bytes, bytes, bytes, bytes]:
return (
_digest(seed, nodeid.split("::", 1)[0]),
_digest(seed, group or nodeid),
_digest(seed, inner_group or group or nodeid),
_digest(seed, nodeid),
)
def _shard_files(path: Path | None) -> frozenset[str] | None:
if path is None:
return None
try:
files: Final = tuple(path.read_text().splitlines())
except (OSError, UnicodeError) as error:
raise pytest.UsageError(f"Cannot read integration shard selection: {error}") from error
if len(frozenset(files)) != len(files) or any(not file for file in files):
raise pytest.UsageError("Integration shard selection contains duplicate or blank files")
return frozenset(files)
@pytest.hookimpl(wrapper=True)
def pytest_collection_modifyitems(config: pytest.Config, items: list[pytest.Item]) -> Iterator[None]:
if config.getoption("dist", default="no") == "loadgroup" or config.getoption("loadgroup", default=False):
automatic_groups: Final = scheduling_groups(tuple(collected_case(item) for item in items))
for item in items:
fixture_group: Final = automatic_groups[item.nodeid]
if fixture_group and not tuple(item.iter_markers("xdist_group")):
item.add_marker(
pytest.mark.xdist_group("integration_fixture_" + hashlib.sha256(fixture_group.encode()).hexdigest())
)
yield
order_seed: Final = INTEGER_OPTION.validate_python(config.getoption("integration_order_seed"))
if order_seed:
cases: Final = tuple(collected_case(item) for item in items)
groups: Final = scheduling_groups(cases)
inner_groups: Final = fixture_groups(cases)
items.sort(
key=lambda item: _order_key(
order_seed,
_case_identity(item.nodeid),
_case_identity(groups[item.nodeid]),
_case_identity(inner_groups[item.nodeid]),
)
)
root: Final = Path(__file__).parent
owned: Final = tuple(
item
for item in items
if item.path.is_relative_to(root) and item.path.relative_to(root).parts[0] in OWNED_DIRECTORIES
)
circleci_only: Final = tuple(
item for item in owned if item.path.relative_to(root.parents[1]).as_posix() not in GITHUB_FILES
)
if circleci_only and os.environ.get("GITHUB_ACTIONS") == "true":
raise pytest.UsageError("Integration contracts are owned by CircleCI")
for item in owned:
item.add_marker(pytest.mark.integration)
config.stash[INVENTORY] = tuple(_case_identity(item.nodeid) for item in owned)
count: Final = INTEGER_OPTION.validate_python(config.getoption("integration_shard_count"))
index: Final = INTEGER_OPTION.validate_python(config.getoption("integration_shard_index"))
if count < 1 or not 0 <= index < count:
raise pytest.UsageError("Integration shard index must be within the positive shard count")
files: Final = config.stash[SHARD_FILES]
if count > 1 and files is None:
raise pytest.UsageError("Parallel integration runs require CircleCI's shard file selection")
if files is not None:
available: Final = frozenset(item.nodeid.split("::", 1)[0] for item in owned)
if files - available:
raise pytest.UsageError(f"Integration shard selected uncollected files: {sorted(files - available)}")
selected: Final = tuple(item for item in items if item not in owned or item.nodeid.split("::", 1)[0] in files)
deselected: Final = tuple(item for item in owned if item.nodeid.split("::", 1)[0] not in files)
config.hook.pytest_deselected(items=deselected)
items[:] = selected
config.stash[COLLECTED] = tuple(_case_identity(item.nodeid) for item in items if item in owned)
def pytest_sessionfinish(session: pytest.Session, exitstatus: int) -> None:
if hasattr(session.config, "workerinput"):
return
destination: Final = os.environ.get("INTEGRATION_RESULTS_DIR")
if destination is None:
return
collected: Final = session.config.stash.get(COLLECTED, ())
reports: Final = tuple(report for report in session.config.stash[REPORTS] if report.nodeid in collected)
inventory: Final = session.config.stash.get(INVENTORY, collected)
empty_shard: Final = session.config.stash[SHARD_FILES] == frozenset() and bool(inventory)
passed: Final = tuple(report.nodeid for report in reports if report.when == "call" and report.passed)
skipped: Final = tuple(report.nodeid for report in reports if report.skipped)
complete: Final = (
(exitstatus == 0 or (empty_shard and exitstatus == pytest.ExitCode.NO_TESTS_COLLECTED))
and (bool(collected) or empty_shard)
and sorted(collected) == sorted(passed + skipped)
and not any(report.failed for report in reports)
)
output: Final = Path(destination)
output.mkdir(parents=True, exist_ok=True)
(output / "execution.json").write_text(
json.dumps(
{
"collected": collected,
"inventory": inventory,
"shard_count": session.config.getoption("integration_shard_count"),
"shard_index": session.config.getoption("integration_shard_index"),
"passed": passed,
"skipped": skipped,
"complete": complete,
"exitstatus": 0 if complete else exitstatus,
"hypothesis_version": version("hypothesis"),
"hypothesis_seed": session.config.getoption("hypothesis_seed"),
"order_seed": session.config.getoption("integration_order_seed"),
"generation": {
"max_examples": LIFECYCLE_SETTINGS.max_examples,
"stateful_step_count": LIFECYCLE_SETTINGS.stateful_step_count,
"database": str(LIFECYCLE_SETTINGS.database),
"phases": [phase.name for phase in LIFECYCLE_SETTINGS.phases],
},
},
indent=2,
)
+ "\n"
)
if not complete and exitstatus == 0:
session.exitstatus = pytest.ExitCode.TESTS_FAILED
if complete and empty_shard:
session.exitstatus = pytest.ExitCode.OK
@pytest.fixture
def gateway() -> Iterator[Gateway]:
with gateway_from_environment() as value:
yield value
@pytest.fixture(scope="session")
def session_gateway() -> Iterator[Gateway]:
with gateway_from_environment() as value:
yield value
@pytest.fixture
def scenario(gateway: Gateway) -> Iterator[Scenario]:
value: Final[Scenario]
with gateway.scenario() as value:
yield value
@pytest_asyncio.fixture(scope="session", loop_scope="session")
async def actors(
session_gateway: Gateway, db: ManagementDatabase
) -> AsyncIterator[ActorSet]:
with session_gateway.scenario() as scenario:
from tests.integration._support.actors import create_actor_set
value: Final[ActorSet] = await create_actor_set(session_gateway, scenario, db)
yield value
@pytest.fixture(scope="session")
def gateway_client(session_gateway: Gateway, actors: ActorSet) -> GatewayClient:
from tests.integration._support.actors import create_gateway_client
return create_gateway_client(session_gateway, actors)
@pytest_asyncio.fixture(scope="session", loop_scope="session")
async def db(session_gateway: Gateway) -> AsyncIterator[ManagementDatabase]:
from tests.integration._support.actors import management_database
value: Final[ManagementDatabase]
async with management_database(session_gateway) as value:
yield value
@pytest_asyncio.fixture(scope="function", loop_scope="session")
async def resources(db: ManagementDatabase) -> AsyncIterator[ScratchResources]:
from tests.integration._support.actors import scratch_resources
value: Final[ScratchResources]
async with scratch_resources(db) as value:
yield value
@pytest.fixture(scope="session")
def shared_provider_server() -> Iterator[SharedProvider]:
if os.environ.get("PYTEST_XDIST_WORKER") and os.environ.get("INTEGRATION_WORKER_ISOLATION") != "1":
pytest.fail("the shared fake provider needs tests to run one at a time; this group runs under pytest-xdist")
with shared_provider() as server:
yield server
late: Final = server.received()
assert late == (), (
f"the shared fake provider got {[item.target for item in late]} after {server.last_test} finished"
)
@pytest.fixture
def provider(shared_provider_server: SharedProvider, request: pytest.FixtureRequest) -> Iterator[SharedProvider]:
stray: Final = shared_provider_server.received()
shared_provider_server.replies.clear()
assert stray == (), (
f"the shared fake provider got {[item.target for item in stray]} "
f"after {shared_provider_server.last_test} finished"
)
yield shared_provider_server
shared_provider_server.last_test = request.node.nodeid
unused: Final = len(shared_provider_server.replies)
unread: Final = shared_provider_server.received()
shared_provider_server.replies.clear()
assert unused == 0, f"{unused} queued provider replies were never requested"
assert unread == (), f"the test never read the provider requests {[item.target for item in unread]}"
@pytest.fixture
def peer(gateway: Gateway) -> Iterator[Gateway]:
url: Final = os.environ["INTEGRATION_PEER_URL"]
assert url.rstrip("/") != str(gateway.client.base_url).rstrip("/")
with Redis(host=os.environ["REDIS_HOST"], port=int(os.environ["REDIS_PORT"])) as cache:
eventually(lambda: cache.pubsub_numsub("litellm_proxy.auth_cache_invalidation")[0][1], lambda count: count >= 2)
with httpx.Client(base_url=url, timeout=15, trust_env=False) as client:
yield Gateway(client, gateway.key, gateway.upstream_url)