[
https://issues.apache.org/jira/browse/BEAM-12324?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17342955#comment-17342955
]
Yifan Mai commented on BEAM-12324:
----------------------------------
Actually we already don't pack combiners when in streaming:
https://github.com/apache/beam/blob/35881d7553335a6c11119b3ad799529d56b51fbe/sdks/python/apache_beam/runners/dataflow/dataflow_runner.py#L467
I wonder if this could be a FnApi thing? In some other logs I am seeing this
which might be relevant:
https://gist.github.com/yifanmai/7ad381a89e99f961fb797f6793fb769a
File "apache_beam/coders/stream.pyx", line 195, in
apache_beam.coders.stream.InputStream.read_var_int64
RuntimeError: VarLong too long.
> TranslationsTest.test_run_packable_combine_* failing on
> PostCommit_Py_VR_Dataflow
> ---------------------------------------------------------------------------------
>
> Key: BEAM-12324
> URL: https://issues.apache.org/jira/browse/BEAM-12324
> Project: Beam
> Issue Type: Bug
> Components: runner-dataflow, sdk-py-core
> Reporter: Brian Hulette
> Assignee: Yifan Mai
> Priority: P1
>
> Failing tests:
> -
> apache_beam.runners.portability.fn_api_runner.translations_test.TranslationsTest.test_run_packable_combine_globally
> -
> apache_beam.runners.portability.fn_api_runner.translations_test.TranslationsTest.test_run_packable_combine_per_key
> First failure:
> https://ci-beam.apache.org/job/beam_PostCommit_Py_VR_Dataflow/7899/
> However note that this suite was timing out since ~April 23 (BEAM-12250), so
> the breakage could have been introduced a while ago.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
