Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix concurrency assumptions for test_worker_reconnects_mid_compute #5623

Merged
merged 3 commits into from
Jan 11, 2022
Merged

Fix concurrency assumptions for test_worker_reconnects_mid_compute #5623

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
5c1b3db
Fix concurrency assumptions for test_worker_reconnects_mid_compute
fjetter Dec 21, 2021
02eeb72
await f1 to control the release transition below
fjetter Dec 21, 2021
8c17054
Update distributed/tests/test_worker.py
fjetter Jan 10, 2022
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
38 changes: 33 additions & 5 deletions distributed/tests/test_worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2411,7 +2411,13 @@ async def test_hold_on_to_replicas(c, s, *workers):
await asyncio.sleep(0.01)


@gen_cluster(client=True)
@gen_cluster(
client=True,
nthreads=[
("", 1),
("", 1),
],
)
async def test_worker_reconnects_mid_compute(c, s, a, b):
"""Ensure that, if a worker disconnects while computing a result, the scheduler will
still accept the result.
Expand Down Expand Up @@ -2479,7 +2485,13 @@ def fast_on_a(lock):
await asyncio.sleep(0.001)


@gen_cluster(client=True)
@gen_cluster(
client=True,
nthreads=[
("", 1),
("", 1),
],
)
async def test_worker_reconnects_mid_compute_multiple_states_on_scheduler(c, s, a, b):
"""
Ensure that a reconnecting worker does not break the scheduler regardless of
Expand All @@ -2494,6 +2506,7 @@ async def test_worker_reconnects_mid_compute_multiple_states_on_scheduler(c, s,
# different states
f1 = c.submit(inc, 1, workers=[a.address], allow_other_workers=True)
f2 = c.submit(inc, f1, workers=[a.address], allow_other_workers=True)
await f1
a_address = a.address

a.periodic_callbacks["heartbeat"].stop()
Expand Down Expand Up @@ -2522,14 +2535,29 @@ def fast_on_a(lock):
while f3.key not in a.tasks:
await asyncio.sleep(0.01)

story_before = s.story(f1.key)
await s.stream_comms[a.address].close()
# Release f1 which triggers a release cycle of all tasks such that
# they are rescheduled on B. However, at this time, B will never be
# able to compute f3 / fast_on_a since it is locked on that worker.
# The only way to get f3 to complete is for Worker A to reconnect.

f1.release()
assert len(s.workers) == 1
while s.tasks[f1.key].state != "released":
await asyncio.sleep(0)
story = s.story(f1.key)
while len(story) == len(story_before):
story = s.story(f1.key)
await asyncio.sleep(0.1)

next = story[len(story_before)]
assert next[:3] == (f1.key, "memory", "released")

a.heartbeat_active = False
await a.heartbeat()
assert len(s.workers) == 2

while len(s.workers) != 2:
await asyncio.sleep(0.01)

# Since B is locked, this is ensured to originate from A
await f3

Expand Down