litellm/tests/unit/integration_support/test_process.py
devin-ai-integration[bot] 7efce7f8e0
test(integration): give chaos and timing tests a harness API and gate raw signals (#46031)
* test(integration): give chaos and timing tests a harness API and gate raw signals

Chaos tests killed and suspended proxy workers with raw os.kill and psutil
calls and bounded elapsed time with fixed literals, so they raced the uvicorn
worker healthcheck and failed on slow CI boxes

OwnedProxy now offers kill_worker and replacing_worker, which return only once
a replacement answers readiness with its own pid, and suspended, which is
refused unless the proxy was launched with a lenient worker healthcheck.
timing.py bounds durations against a measured control or the injected delay.
run.py measures the proxy import time once and derives readiness deadlines and
the per-test timeout from it, writing boot-calibration.json

check_test_quality.py adds TQ010 (raw process signals) and TQ011 (elapsed time
asserted under a numeric literal) for tests/integration outside _support, and
every call site is migrated so both report zero with no suppressions

* test(integration): steer the assistants wire burst onto both workers

The mixed-status burst assumed the OS would spread 48 connections across both workers and failed when one worker took them all. Each half is now sent while the other worker is suspended, and the test asserts an exact 24 and 24 split

* test(integration): restore product-enforced timing bounds lost in the migration

Some TQ011 migrations dropped or loosened a bound whose time is the behaviour
under test. The websocket session cap only kept its floor, so a cap closing at
300 s would have passed, and several timeouts were bounded by the injected stall
instead of the configured product value

assert_enforced_within bounds an elapsed time by the configured value plus
explicit slack, with an optional floor. The websocket cap, the MCP validation
timeout, the guardrail litellm_params timeout, the prisma command timeouts, the
baseline projection lock timeout and the key metadata statement timeout use it.
The device code refusal keeps its 15 s margin, and the langfuse retried prompt
fetch is now bounded by a cold fetch of a steady prompt as well as by the
ignored retry-after

* fix(tests): wait for the supervisor to reap killed workers and fix real-kill migrations

The lenient worker healthcheck was the whole case budget, so uvicorn's supervisor sat
in a ping on the killed worker for up to that long before respawning it. It now uses
the graceful stop budget, and a worker kill returns only after the victim is reaped,
so pooled connections no longer race the dead worker's socket.

Several tests on main picked the multiprocessing resource tracker as their victim and
never killed a worker. Now that they kill a real one, presidio counts streams cut by
the death, delete_model waits for the config reload that removes a deleted model from
a replacement worker, and the wedged coordination Redis test kills before the wedge
and waits for every parked publish to fail before resuming Redis.

The redis cluster test imports the harness through the tests package so it collects
in the unit job, and the per-model fallback rig exposes the owned proxy it now needs.

* test(observability): only require S3 deliveries the killed worker could not have buffered

A real SIGKILL drops the victim's unflushed batch, so requests it already answered can't be expected to land. The test now requires every request served after the replacement to land, some of the surviving worker's burst share to land, and nothing unsent to land

---------

Co-authored-by: Yuneng Jiang <yuneng@berri.ai>
2026-10-12 05:27:33 -07:00

450 lines
18 KiB
Python

from __future__ import annotations
import asyncio
import errno
import importlib
import os
import socket
import subprocess
import sys
import threading
import time
from collections.abc import Iterator
from contextlib import contextmanager
from pathlib import Path
from types import ModuleType
from typing import Final
from unittest.mock import MagicMock
import httpx
import psutil
import pytest
import yaml
from pydantic import TypeAdapter
from tests.integration._support.client import Gateway, eventually
from tests.integration._support.process import (
OwnedProxy,
WorkerHealthcheck,
WorkerKill,
healthcheck_seconds,
kill_stray_process,
owned_proxy_process,
replacement_pid,
stop_process_group,
worker_configuration,
)
TESTS_DIR: Final = Path(__file__).resolve().parents[2]
UNRELATED_CRASH: Final = "Traceback (most recent call last):\nModuleNotFoundError: No module named 'litellm'\n"
def test_worker_configuration_routes_stock_edges_to_its_own_servers_and_preserves_other_targets(
monkeypatch: pytest.MonkeyPatch, tmp_path: Path
) -> None:
monkeypatch.setenv("INTEGRATION_WORKER_ISOLATION", "1")
monkeypatch.setenv("INTEGRATION_PROVIDER_PORT", "8203")
monkeypatch.setenv("INTEGRATION_UPSTREAM_URL", "http://127.0.0.1:8202")
source: Final = tmp_path / "source.yaml"
original: Final = (
'api_base: http://127.0.0.1:8191/v1\ntoken_url: "http://127.0.0.1:8190/_oauth/token"\n'
"peer: http://127.0.0.1:81910/v1\ncustom: http://127.0.0.1:9000/v1\n"
)
source.write_text(original)
configured: Final = worker_configuration(source, tmp_path / "worker")
assert yaml.safe_load(configured.read_text()) == {
"api_base": "http://127.0.0.1:8203/v1",
"token_url": "http://127.0.0.1:8202/_oauth/token",
"peer": "http://127.0.0.1:81910/v1",
"custom": "http://127.0.0.1:9000/v1",
"general_settings": {"database_connection_pool_limit": 2},
}
assert source.read_text() == original
@pytest.mark.parametrize("limit", (1, 5, 0, None, "", "bad"))
def test_worker_configuration_preserves_explicit_pool_limits(
monkeypatch: pytest.MonkeyPatch, tmp_path: Path, limit: int | str | None
) -> None:
monkeypatch.setenv("INTEGRATION_WORKER_ISOLATION", "1")
source: Final = tmp_path / "source.yaml"
original: Final = {
"general_settings": {"database_connection_pool_limit": limit, "database_connection_timeout": 3},
"model_list": [{"model_name": "synthetic", "litellm_params": {"model": "openai/synthetic"}}],
}
source.write_text(yaml.safe_dump(original))
configured: Final = worker_configuration(source, tmp_path / "worker")
assert yaml.safe_load(configured.read_text()) == original
assert yaml.safe_load(source.read_text()) == original
def test_worker_configuration_budgets_implicit_pools_without_changing_connection_options(
monkeypatch: pytest.MonkeyPatch, tmp_path: Path
) -> None:
monkeypatch.setenv("INTEGRATION_WORKER_ISOLATION", "1")
source: Final = tmp_path / "source.yaml"
original: Final = {"general_settings": {"database_extra_connection_params": {"connection_limit": "1"}}}
source.write_text(yaml.safe_dump(original))
configured: Final = worker_configuration(source, tmp_path / "worker")
decoded: Final = TypeAdapter(dict[str, object]).validate_python(yaml.safe_load(configured.read_text()))
assert decoded == {
"general_settings": {
"database_connection_pool_limit": 2,
"database_extra_connection_params": {"connection_limit": "1"},
}
}
assert yaml.safe_load(source.read_text()) == original
def test_serial_proxy_keeps_the_requested_configuration(monkeypatch: pytest.MonkeyPatch, tmp_path: Path) -> None:
monkeypatch.delenv("INTEGRATION_WORKER_ISOLATION", raising=False)
source: Final = tmp_path / "source.yaml"
source.write_text("api_base: http://127.0.0.1:8191/v1\n")
assert worker_configuration(source, tmp_path / "unused") == source
assert source.read_text() == "api_base: http://127.0.0.1:8191/v1\n"
@pytest.fixture
def process_module(monkeypatch: pytest.MonkeyPatch) -> ModuleType:
monkeypatch.syspath_prepend(str(TESTS_DIR))
return importlib.import_module("integration._support.process")
async def _refused_bind(port: int) -> OSError | None:
try:
await asyncio.get_running_loop().create_server(asyncio.Protocol, "127.0.0.1", port)
except OSError as refused:
return refused
return None
@pytest.fixture
def bind_error_line() -> Iterator[str]:
with socket.socket() as held:
held.bind(("127.0.0.1", 0))
held.listen()
refused: Final = asyncio.run(_refused_bind(held.getsockname()[1]))
assert refused is not None and refused.errno == errno.EADDRINUSE
yield f"ERROR: {refused}\n"
def _written_log(directory: Path, text: str) -> Path:
log: Final = directory / "owned-proxy.log"
log.write_text(text)
return log
def test_lost_port_race_matches_the_bind_error_the_server_logs(
process_module: ModuleType, bind_error_line: str, tmp_path: Path
) -> None:
assert process_module._lost_port_race(1, _written_log(tmp_path, bind_error_line))
def test_lost_port_race_ignores_an_exit_for_another_reason(process_module: ModuleType, tmp_path: Path) -> None:
assert not process_module._lost_port_race(1, _written_log(tmp_path, UNRELATED_CRASH))
def test_lost_port_race_needs_the_process_to_have_exited(
process_module: ModuleType, bind_error_line: str, tmp_path: Path
) -> None:
assert not process_module._lost_port_race(None, _written_log(tmp_path, bind_error_line))
@pytest.mark.parametrize("failure", (psutil.AccessDenied, psutil.NoSuchProcess))
def test_an_unreadable_process_cannot_establish_listener_ownership(
process_module: ModuleType, monkeypatch: pytest.MonkeyPatch, tmp_path: Path, failure: type[psutil.Error]
) -> None:
member: Final = MagicMock(spec=psutil.Process)
member.pid = os.getpid()
member.net_connections.side_effect = failure(member.pid)
monkeypatch.setattr(psutil, "process_iter", lambda: iter((member,)))
parent: Final = MagicMock()
parent.pid = os.getpgid(member.pid)
launch: Final = process_module._Launch(parent, 8123, tmp_path / "unused.log", 0.0)
assert not process_module._owns_listener(launch)
def test_an_unreadable_member_does_not_hide_another_owned_listener(
process_module: ModuleType, monkeypatch: pytest.MonkeyPatch, tmp_path: Path
) -> None:
unreadable: Final = MagicMock(spec=psutil.Process)
unreadable.pid = 1
unreadable.net_connections.side_effect = psutil.AccessDenied(1)
readable: Final = MagicMock(spec=psutil.Process)
readable.pid = 2
listener: Final = MagicMock()
listener.status = psutil.CONN_LISTEN
listener.laddr.port = 8123
readable.net_connections.return_value = (listener,)
monkeypatch.setattr(psutil, "process_iter", lambda: iter((unreadable, readable)))
monkeypatch.setattr(os, "getpgid", lambda pid: 99)
parent: Final = MagicMock()
parent.pid = 99
launch: Final = process_module._Launch(parent, 8123, tmp_path / "unused.log", 0.0)
assert process_module._owns_listener(launch)
@pytest.mark.parametrize(
("status", "ready"),
(
(psutil.STATUS_RUNNING, True),
(psutil.STATUS_SLEEPING, True),
(psutil.STATUS_ZOMBIE, False),
(psutil.STATUS_DEAD, False),
(psutil.STATUS_STOPPED, False),
(psutil.STATUS_TRACING_STOP, False),
(psutil.AccessDenied, False),
(psutil.NoSuchProcess, False),
),
)
def test_only_live_readable_workers_can_satisfy_readiness(
process_module: ModuleType, monkeypatch: pytest.MonkeyPatch, tmp_path: Path, status: str | type[psutil.Error], ready: bool
) -> None:
member: Final = MagicMock(spec=psutil.Process)
member.pid = os.getpid()
if isinstance(status, str):
member.status.return_value = status
else:
member.status.side_effect = status(member.pid)
listener: Final = MagicMock()
listener.status = psutil.CONN_LISTEN
listener.laddr.port = 8123
member.net_connections.return_value = (listener,)
monkeypatch.setattr(psutil, "process_iter", lambda: iter((member,)))
parent: Final = MagicMock()
parent.pid = os.getpgid(member.pid)
parent.poll.return_value = None
launch: Final = process_module._Launch(parent, 8123, tmp_path / "unused.log", 0.0)
original_client: Final = httpx.Client
def client(*, base_url: str, timeout: float, trust_env: bool, limits: httpx.Limits) -> httpx.Client:
return original_client(
base_url=base_url,
timeout=timeout,
trust_env=trust_env,
limits=limits,
transport=httpx.MockTransport(
lambda request: httpx.Response(200, headers={"x-integration-worker-pid": str(member.pid)})
),
)
monkeypatch.setattr(httpx, "Client", client)
monkeypatch.setattr(time, "monotonic", lambda: 0.0)
monkeypatch.setenv("INTEGRATION_PROXY_READY_SECONDS", "-1")
if ready:
process_module._wait_until_ready(launch)
else:
with pytest.raises(AssertionError, match="readiness deadline"):
process_module._wait_until_ready(launch)
@pytest.mark.parametrize(
("writer", "explicit_reader", "expected_reader"),
(
("postgresql://writer@127.0.0.1/shared", None, "postgresql://reader@127.0.0.1/shared"),
("postgresql://writer@127.0.0.1/scratch", None, None),
(
"postgresql://writer@127.0.0.1/scratch",
"postgresql://reader@127.0.0.1/scratch",
"postgresql://reader@127.0.0.1/scratch",
),
),
)
def test_owned_writer_keeps_only_a_reader_paired_with_its_database(
process_module: ModuleType,
monkeypatch: pytest.MonkeyPatch,
writer: str,
explicit_reader: str | None,
expected_reader: str | None,
) -> None:
monkeypatch.setenv("INTEGRATION_PROXY_READ_REPLICA_URL", "postgresql://reader@127.0.0.1/shared")
overrides: Final = {
"DATABASE_URL": writer,
**({"DATABASE_URL_READ_REPLICA": explicit_reader} if explicit_reader else {}),
}
with httpx.Client(base_url="http://127.0.0.1:1", trust_env=False) as client:
environment: Final = process_module._proxy_environment(
Gateway(client, "owned-test-key", "http://127.0.0.1:1"), overrides, ()
)
assert environment["DATABASE_URL"] == writer
assert environment.get("DATABASE_URL_READ_REPLICA") == expected_reader
@pytest.mark.parametrize("explicit", (False, True))
@pytest.mark.parametrize("remove", (False, True))
def test_owned_runtime_configuration_preserves_custom_worker_startup_hooks(
process_module: ModuleType, monkeypatch: pytest.MonkeyPatch, explicit: bool, remove: bool
) -> None:
monkeypatch.setenv("LITELLM_WORKER_STARTUP_HOOKS", "ambient:configure")
with httpx.Client(base_url="http://127.0.0.1:1", trust_env=False) as client:
environment: Final = process_module._proxy_environment(
Gateway(client, "owned-test-key", "http://127.0.0.1:1"),
{"LITELLM_WORKER_STARTUP_HOOKS": "custom:configure"} if explicit else {},
("LITELLM_WORKER_STARTUP_HOOKS",) if remove else (),
)
expected: Final = "custom:configure" if explicit else ("" if remove else "ambient:configure")
assert environment["LITELLM_WORKER_STARTUP_HOOKS"] == ",".join(
hook
for hook in (
"integration._support.runtime:configure_executor",
"integration._support.runtime:configure_readiness",
expected,
)
if hook
)
_SUPERVISOR: Final = """import multiprocessing
import time
if __name__ == "__main__":
worker = multiprocessing.get_context("spawn").Process(target=time.sleep, args=(60,))
worker.start()
print(worker.pid, flush=True)
deadline = time.monotonic() + 60
while time.monotonic() < deadline:
multiprocessing.active_children()
time.sleep(0.05)
"""
@contextmanager
def _supervised(tmp_path: Path, healthcheck: WorkerHealthcheck) -> Iterator[tuple[OwnedProxy, int]]:
script: Final = tmp_path / "supervisor.py"
script.write_text(_SUPERVISOR)
with subprocess.Popen(
(sys.executable, "-P", str(script)), stdout=subprocess.PIPE, start_new_session=True
) as supervisor:
try:
assert supervisor.stdout is not None
worker: Final = int(supervisor.stdout.readline())
with httpx.Client(base_url="http://127.0.0.1:9") as client:
owned: Final = OwnedProxy(
Gateway(client, "sk-unit", "http://127.0.0.1:9"), supervisor, tmp_path / "proxy.log", healthcheck
)
eventually(owned.worker_pids, lambda pids: pids == frozenset((worker,)), seconds=30)
yield owned, worker
finally:
stop_process_group(supervisor, 5)
def test_worker_pids_are_the_supervisors_spawned_workers_and_not_the_supervisor(tmp_path: Path) -> None:
with _supervised(tmp_path, WorkerHealthcheck.STRICT) as (owned, worker):
assert owned.worker_pids() == frozenset((worker,))
assert owned.process.pid not in owned.worker_pids()
def test_only_a_serving_worker_of_this_proxy_can_be_killed(tmp_path: Path) -> None:
with _supervised(tmp_path, WorkerHealthcheck.STRICT) as (owned, worker):
with pytest.raises(AssertionError, match="not a serving worker"):
owned.kill_worker(owned.process.pid)
assert owned.process.poll() is None
assert psutil.Process(worker).is_running()
def _leave_without_killing(owned: OwnedProxy, worker: int) -> None:
with owned.replacing_worker(worker):
pass
def test_a_replacing_block_that_never_kills_its_worker_fails_and_leaves_it_serving(tmp_path: Path) -> None:
with _supervised(tmp_path, WorkerHealthcheck.STRICT) as (owned, worker):
with pytest.raises(AssertionError, match="without killing the worker"):
_leave_without_killing(owned, worker)
assert owned.worker_pids() == frozenset((worker,))
def _kill_of(owned: OwnedProxy, worker: int) -> WorkerKill:
return WorkerKill(psutil.Process(worker), psutil.Process(owned.process.pid), 5, threading.Event())
def test_a_worker_kill_returns_once_the_supervisor_reaped_the_victim_and_cannot_fire_twice(tmp_path: Path) -> None:
with _supervised(tmp_path, WorkerHealthcheck.STRICT) as (owned, worker):
kill: Final = _kill_of(owned, worker)
kill()
assert not psutil.pid_exists(worker)
assert kill.fired.is_set()
with pytest.raises(AssertionError, match="already killed"):
kill()
def test_a_frozen_victim_stays_stopped_until_the_kill(tmp_path: Path) -> None:
with _supervised(tmp_path, WorkerHealthcheck.STRICT) as (owned, worker):
kill: Final = _kill_of(owned, worker)
assert kill.freeze().status() == psutil.STATUS_STOPPED
kill()
assert not psutil.pid_exists(worker)
def test_a_kill_under_a_suspended_supervisor_returns_with_the_victim_left_unreaped(tmp_path: Path) -> None:
with _supervised(tmp_path, WorkerHealthcheck.LENIENT) as (owned, worker):
kill: Final = _kill_of(owned, worker)
with owned.suspended(owned.process.pid):
kill()
assert psutil.Process(worker).status() == psutil.STATUS_ZOMBIE
eventually(lambda: psutil.pid_exists(worker), lambda exists: not exists, seconds=5)
def test_a_kill_fails_when_the_supervisor_never_reaps_the_victim(tmp_path: Path) -> None:
with _supervised(tmp_path, WorkerHealthcheck.STRICT) as (owned, worker):
unreaping: Final = psutil.Process(owned.process.pid)
kill: Final = WorkerKill(psutil.Process(worker), psutil.Process(os.getpid()), 0.5, threading.Event())
unreaping.suspend()
try:
with pytest.raises(AssertionError, match="did not reap"):
kill()
finally:
unreaping.resume()
assert not kill.fired.is_set()
def test_a_strict_healthcheck_proxy_refuses_suspension_before_stopping_anything(tmp_path: Path) -> None:
with _supervised(tmp_path, WorkerHealthcheck.STRICT) as (owned, worker):
with pytest.raises(AssertionError, match="LENIENT"):
with owned.suspended(worker):
pass
assert psutil.Process(worker).status() != psutil.STATUS_STOPPED
def test_a_lenient_healthcheck_proxy_suspends_for_the_block_and_resumes_after(tmp_path: Path) -> None:
with _supervised(tmp_path, WorkerHealthcheck.LENIENT) as (owned, worker):
with owned.suspended(worker) as suspended:
assert suspended.status() == psutil.STATUS_STOPPED
assert psutil.Process(worker).status() != psutil.STATUS_STOPPED
assert owned.worker_pids() == frozenset((worker,))
def test_a_worker_killed_while_suspended_fails_the_block(tmp_path: Path) -> None:
with _supervised(tmp_path, WorkerHealthcheck.LENIENT) as (owned, worker):
with pytest.raises(AssertionError, match="died while suspended"):
with owned.suspended(worker):
kill_stray_process(worker)
def test_the_worker_healthcheck_cannot_be_overridden_through_extra_arguments(tmp_path: Path) -> None:
with pytest.raises(AssertionError, match="healthcheck"):
with owned_proxy_process(
MagicMock(spec=Gateway), tmp_path, {}, extra_arguments=("--timeout_worker_healthcheck", "60")
):
pass
assert not tuple(tmp_path.glob("owned-proxy-*.log"))
def test_a_lenient_healthcheck_outlasts_a_strict_one() -> None:
assert healthcheck_seconds(WorkerHealthcheck.LENIENT) > healthcheck_seconds(WorkerHealthcheck.STRICT)
@pytest.mark.parametrize(
("answered", "serving", "replacement"),
(
(30, frozenset((20, 30)), 30),
(20, frozenset((20, 30)), None),
(10, frozenset((20, 30)), None),
(40, frozenset((20, 30)), None),
(None, frozenset((20, 30)), None),
),
)
def test_only_a_new_serving_worker_answering_readiness_counts_as_the_replacement(
answered: int | None, serving: frozenset[int], replacement: int | None
) -> None:
assert replacement_pid(frozenset((10, 20)), answered, serving) == replacement