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

Ensure reconnecting workers do not loose required data #5436

Merged
merged 3 commits into from
Oct 18, 2021
Merged

Ensure reconnecting workers do not loose required data #5436

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2545768
Ensure reconnecting workers do not loose required data
fjetter Oct 18, 2021
19d4bb8
fix free_keys calls
fjetter Oct 18, 2021
f9bb38f
attach timestamp to stim ID of delete_worker_data
fjetter Oct 18, 2021
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: 13 additions & 9 deletions distributed/scheduler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2737,8 +2737,8 @@ def transition_processing_memory(

if ws != ts._processing_on: # someone else has this task
logger.info(
"Unexpected worker completed task, likely due to "
"work stealing. Expected: %s, Got: %s, Key: %s",
"Unexpected worker completed task %s. Expected: %s, Got: %s, Key: %s",
ts.key,
ts._processing_on,
ws,
key,
Expand Down Expand Up @@ -2835,7 +2835,7 @@ def transition_memory_released(self, key, safe: bint = False):
worker_msg = {
"op": "free-keys",
"keys": [key],
"reason": f"Memory->Released {key}",
"stimulus_id": f"memory-released-{time()}",
}
for ws in ts._who_has:
worker_msgs[ws._address] = [worker_msg]
Expand Down Expand Up @@ -2935,7 +2935,11 @@ def transition_erred_released(self, key):
if dts._state == "erred":
recommendations[dts._key] = "waiting"

w_msg = {"op": "free-keys", "keys": [key], "reason": "Erred->Released"}
w_msg = {
"op": "free-keys",
"keys": [key],
"stimulus_id": f"erred-released-{time()}",
}
for ws_addr in ts._erred_on:
worker_msgs[ws_addr] = [w_msg]
ts._erred_on.clear()
Expand Down Expand Up @@ -3013,7 +3017,7 @@ def transition_processing_released(self, key):
{
"op": "free-keys",
"keys": [key],
"reason": f"processing-released-{time()}",
"stimulus_id": f"processing-released-{time()}",
}
]

Expand Down Expand Up @@ -4339,9 +4343,9 @@ async def add_worker(
worker_msgs[address] = []
worker_msgs[address].append(
{
"op": "free-keys",
"op": "remove-replicas",
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is the relevant change

"keys": already_released_keys,
"reason": f"reconnect-already-released-{time()}",
"stimulus_id": f"reconnect-already-released-{time()}",
}
)
for ts in list(parent._unrunnable):
Expand Down Expand Up @@ -4767,7 +4771,7 @@ def stimulus_task_finished(self, key=None, worker=None, **kwargs):
{
"op": "free-keys",
"keys": [key],
"reason": f"already-released-or-forgotten-{time()}",
"stimulus_id": f"already-released-or-forgotten-{time()}",
}
]
elif ts._state == "memory":
Expand Down Expand Up @@ -7846,7 +7850,7 @@ def _propagate_forgotten(
{
"op": "free-keys",
"keys": [key],
"reason": f"propagate-forgotten {ts.key}",
"stimulus_id": f"propagate-forgotten-{time()}",
}
]
state.remove_all_replicas(ts)
Expand Down
2 changes: 0 additions & 2 deletions distributed/tests/test_worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2444,7 +2444,6 @@ async def test_hold_on_to_replicas(c, s, *workers):
await asyncio.sleep(0.01)


@pytest.mark.flaky(reruns=10, reruns_delay=5)
@gen_cluster(client=True)
async def test_worker_reconnects_mid_compute(c, s, a, b):
"""Ensure that, if a worker disconnects while computing a result, the scheduler will
Expand Down Expand Up @@ -2513,7 +2512,6 @@ def fast_on_a(lock):
await asyncio.sleep(0.001)


@pytest.mark.flaky(reruns=10, reruns_delay=5)
@gen_cluster(client=True)
async def test_worker_reconnects_mid_compute_multiple_states_on_scheduler(c, s, a, b):
"""
Expand Down
6 changes: 3 additions & 3 deletions distributed/worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1596,7 +1596,7 @@ def update_data(
self.batched_stream.send(msg)
return {"nbytes": {k: sizeof(v) for k, v in data.items()}, "status": "OK"}

def handle_free_keys(self, comm=None, keys=None, reason=None):
def handle_free_keys(self, comm=None, keys=None, stimulus_id=None):
"""
Handler to be called by the scheduler.

Expand All @@ -1607,14 +1607,14 @@ def handle_free_keys(self, comm=None, keys=None, reason=None):
still decide to hold on to the data and task since it is required by an
upstream dependency.
"""
self.log.append(("free-keys", keys, reason))
self.log.append(("free-keys", keys, stimulus_id))
recommendations = {}
for key in keys:
ts = self.tasks.get(key)
if ts:
recommendations[ts] = "released" if ts.dependents else "forgotten"

self.transitions(recommendations, stimulus_id=reason)
self.transitions(recommendations, stimulus_id=stimulus_id)

def handle_remove_replicas(self, keys, stimulus_id):
"""Stream handler notifying the worker that it might be holding unreferenced,
Expand Down