[
https://issues.apache.org/jira/browse/BEAM-14476?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17539835#comment-17539835
]
Deepak Nagaraj commented on BEAM-14476:
---------------------------------------
[~tvalentyn] I have attached a complete pipeline. It does not finish processing
all the elements. The worker pool threads seem to get stuck in a deadlock. I
invoke it as follows (I have a Flink cluster running on my laptop, along with
Beam job server and expansion service).
{code:java}
python pipeline_min.py submit --job_server=localhost {code}
You may have to tweak the number of elements and the print() size to suit your
environment. It essentially seems to be a variant of Python bug 42717.
https://bugs.python.org/issue42717
> Verbose print causes pipelines to get stuck on FlinkRunner.
> -----------------------------------------------------------
>
> Key: BEAM-14476
> URL: https://issues.apache.org/jira/browse/BEAM-14476
> Project: Beam
> Issue Type: Bug
> Components: runner-flink, sdk-py-harness
> Reporter: Robert Bradshaw
> Priority: P2
> Labels: interrupts
> Attachments: pipeline_min.py
>
>
> Minimal reproduction:
>
> {{def _run_pipeline(pipeline):}}
> {{ def process_data(unused):}}
> {{ print('a'*1000)}}
> {{ _ = (}}
> {{ pipeline}}
> {{ | "Create" >> beam.Create(['a']*1000)}}
> {{ | "Process" >> beam.Map(process_data)}}
> {{ )}}
> {{ pipeline.run().wait_until_finish()}}
> See thread at https://lists.apache.org/thread/r971f432mvko4g3hmnrodh8s4n2wkk6p
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
