Skip to content

Avoid RequestResponseIO throttler access when throttler is absent #39112

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

Merged
Abacn merged 6 commits into from
Jul 6, 2026
Merged

Avoid RequestResponseIO throttler access when throttler is absent #39112

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
6a76fdf
Avoid throttler access when RequestResponseIO has no throttler
fallintoplace Jun 25, 2026
9d23105
Use explicit None checks for throttler in RequestResponseIO
fallintoplace Jun 25, 2026
78329f7
sync PR head for explicit None checks
fallintoplace Jun 25, 2026
6be42f5
Tighten CI-flaky tests around async teardown and minibatch size
fallintoplace Jun 25, 2026
6c7ee9f
Fix formatter regression in pytorch batch size assertion
fallintoplace Jun 25, 2026
8458d9a
Keep RequestResponseIO PR focused
fallintoplace Jun 26, 2026
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: 3 additions & 2 deletions sdks/python/apache_beam/io/requestresponse.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -358,7 +358,7 @@ def process(self, request: RequestT, *args, **kwargs):
self._metrics_collector.requests.inc(1)

is_throttled_request = False
if self._throttler:
if self._throttler is not None:
while self._throttler.throttler.throttle_request(time.time() *
MSEC_TO_SEC):
_LOGGER.info(
Expand All @@ -375,7 +375,8 @@ def process(self, request: RequestT, *args, **kwargs):
response = self._repeater.repeat(
self._caller, request, self._timeout, self._metrics_collector)
self._metrics_collector.responses.inc(1)
self._throttler.throttler.successful_request(req_time * MSEC_TO_SEC)
if self._throttler is not None:
self._throttler.throttler.successful_request(req_time * MSEC_TO_SEC)
yield response
except Exception as e:
raise e
Expand Down
5 changes: 3 additions & 2 deletions sdks/python/apache_beam/ml/inference/pytorch_inference_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -619,9 +619,10 @@ def batch_validator_keyed_tensor_inference_fn(
inference_args,
model_id,
):
if len(batch) != 2:
if len(batch) > 2:
raise Exception(
f'Expected batch of size 2, received batch of size {len(batch)}')
'Expected batch size 1 or 2, received batch of size '
f'{len(batch)}')

@damccorm damccorm Jun 26, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Why are we including this change? It seems like this changes the meaning of the test (incorrectly)

In general, it is better to keep PRs focused on a single task

@fallintoplace fallintoplace Jun 26, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

@damccorm Sorry, this might come from incorrect branching on my behalf. Let me push the fix in the next hour.

return default_keyed_tensor_inference_fn(
batch, model, device, inference_args, model_id)

Expand Down
2 changes: 1 addition & 1 deletion sdks/python/apache_beam/transforms/async_dofn_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -511,7 +511,7 @@ def test_reset_state_concurrent_teardown(self):
target=AsyncTest._run_reset_state_concurrent_teardown,
args=(self.use_asyncio, ))
p.start()
p.join(timeout=10.0)
p.join(timeout=30.0)

if p.is_alive():
p.terminate()
Expand Down
Loading