Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 2 additions & 3 deletions langfuse/_utils/prompt_cache.py
Original file line number Diff line number Diff line change
Expand Up @@ -134,9 +134,8 @@ def shutdown(self) -> None:

atexit.unregister(self.shutdown)

for consumer in self._consumers:
consumer.pause()

# Busy workers must consume their sentinels after finishing queued work.
# Pausing them here would leave unfinished items in the queue.
for _ in self._consumers:
self._queue.put(_SHUTDOWN_SENTINEL)

Expand Down
45 changes: 41 additions & 4 deletions tests/unit/test_prompt.py
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
import threading
from unittest.mock import Mock, patch

import pytest
Expand Down Expand Up @@ -143,7 +144,7 @@ def wait_for_prompt_refresh(langfuse: Langfuse) -> None:
langfuse._resources.prompt_cache._task_manager.wait_for_idle()


def test_prompt_cache_task_manager_pauses_all_workers_before_broadcasting_shutdown():
def test_prompt_cache_task_manager_broadcasts_shutdown_before_joining_workers():
manager = PromptCacheTaskManager(threads=0)
events = []

Expand All @@ -167,9 +168,6 @@ def put(self, item):
manager.shutdown()

assert [event[0] for event in events] == [
"pause",
"pause",
"pause",
"put",
"put",
"put",
Expand All @@ -179,6 +177,45 @@ def put(self, item):
]


@pytest.mark.parametrize("workers", [1, 3])
@pytest.mark.parametrize("task_fails", [False, True])
def test_prompt_cache_shutdown_drains_busy_workers(workers, task_fails):
manager = PromptCacheTaskManager(threads=workers)
started = threading.Barrier(workers + 1)
release = threading.Event()

def refresh():
started.wait(timeout=5)
release.wait(timeout=5)
if task_fails:
raise RuntimeError("Prompt refresh failed")

shutdown = threading.Thread(target=manager.shutdown, daemon=True)
try:
for index in range(workers):
manager.add_task(str(index), refresh)
started.wait(timeout=5)
shutdown.start()
# Keep all real workers busy until shutdown has queued their sentinels.
with manager._queue.not_empty:
assert manager._queue.not_empty.wait_for(
lambda: len(manager._queue.queue) == workers, timeout=5
)
finally:
release.set()
if shutdown.ident is not None:
shutdown.join(timeout=5)
else:
manager.shutdown()

assert not shutdown.is_alive()
assert all(not consumer.is_alive() for consumer in manager._consumers)
assert manager.active_tasks() == 0
assert manager._queue.unfinished_tasks == 0
# The queue assertion prevents this behavioral check from hanging on regressions.
manager.wait_for_idle()


def test_get_fresh_prompt(langfuse):
prompt_name = "test_get_fresh_prompt"
prompt = Prompt_Text(
Expand Down