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

Jump to bottom

Reduce complexity of test_gather_allow_worker_reconnect #4739

Merged
merged 1 commit into from
May 6, 2021
Merged

Reduce complexity of test_gather_allow_worker_reconnect #4739

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
22 changes: 8 additions & 14 deletions distributed/tests/test_scheduler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2037,7 +2037,6 @@ async def test_gather_no_workers(c, s, a, b):
assert list(res["keys"]) == ["x"]


@pytest.mark.flaky(reruns=10, reruns_delay=5, condition=MACOS)
@gen_cluster(client=True, client_kwargs={"direct_to_workers": False})
async def test_gather_allow_worker_reconnect(c, s, a, b):
"""
Expand All @@ -2056,9 +2055,10 @@ async def test_gather_allow_worker_reconnect(c, s, a, b):
def inc_slow(x):
# Once the graph below is rescheduled this computation runs again. We
# need to sleep for at least 0.5 seconds to give the worker a chance to
# reconnect (Heartbeat timing)
# reconnect (Heartbeat timing). In slow CI situations, the actual
# reconnect might take a bit longer, therefore wait more
if x in already_calculated:
time.sleep(1)
time.sleep(2)
already_calculated.append(x)
return x + 1

Expand All @@ -2070,18 +2070,16 @@ def reducer(x, y):

z = c.submit(reducer, x, y)

s.rpc = await FlakyConnectionPool(failing_connections=4)
s.rpc = await FlakyConnectionPool(failing_connections=1)

with dask.config.set(
{"distributed.comm.retry.delay_min": 0.5, "distributed.comm.retry.count": 3}
):
# This behaviour is independent of retries. Remove them to reduce complexity
# of this setup
with dask.config.set({"distributed.comm.retry.count": 0}):
with captured_logger(
logging.getLogger("distributed.scheduler")
) as sched_logger, captured_logger(
logging.getLogger("distributed.client")
) as client_logger, captured_logger(
logging.getLogger("distributed.utils_comm")
) as utils_comm_logger:
) as client_logger:
# Gather using the client (as an ordinary user would)
# Upon a missing key, the client will reschedule the computations
res = await c.gather(z)
Expand All @@ -2090,14 +2088,11 @@ def reducer(x, y):

sched_logger = sched_logger.getvalue()
client_logger = client_logger.getvalue()
utils_comm_logger = utils_comm_logger.getvalue()

# Ensure that the communication was done via the scheduler, i.e. we actually hit a
# bad connection
assert s.rpc.cnn_count > 0

assert "Retrying get_data_from_worker after exception" in utils_comm_logger

# The reducer task was actually not found upon first collection. The client will
# reschedule the graph
assert "Couldn't gather 1 keys, rescheduling" in client_logger
Expand All @@ -2116,7 +2111,6 @@ def reducer(x, y):
]
assert len(transitions_to_processing) == 1

starts = []
finish_processing_transitions = 0
for transition in s.transition_log:
key, start, finish, recommendations, timestamp = transition
Expand Down