robertwb commented on a change in pull request #11314: [BEAM-9562] Send Timers
over Data Channel as Elements
URL: https://github.com/apache/beam/pull/11314#discussion_r405826837
##########
File path: sdks/python/apache_beam/runners/worker/bundle_processor.py
##########
@@ -846,6 +870,30 @@ def process_bundle(self, instruction_id):
data_channels[input_op.data_channel].append(input_op.transform_id)
input_op_by_transform_id[input_op.transform_id] = input_op
+ # Set up timer output stream
+ timer_output_streams = {}
+ for transform_id, timer_list in self.timer_ids.items():
+ output_streams = {}
+ for timer_id in timer_list:
+ output_streams[
+ timer_id] = self.timer_data_channel.output_timer_stream(
+ instruction_id, transform_id, timer_id)
+ timer_output_streams[transform_id] = output_streams
+ self.process_timer_ops[
+ transform_id].user_state_context.update_timer_output_streams(
+ output_streams)
+
+ # Process timers
+ if self.timer_data_channel:
Review comment:
We can't safely assume the runner will finish sending all timers before
sending any of the data (and the buffer may get full, resulting in a deadlock).
I think we need to have a data_channel.inputs() that returns both data and
timers and then branch in the loop.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
