You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flink.apache.org by "Huang Xingbo (Jira)" <ji...@apache.org> on 2021/04/29 02:11:00 UTC

[jira] [Created] (FLINK-22517) Fix pickle compatibility problem in different Python versions

Huang Xingbo created FLINK-22517:
------------------------------------

             Summary: Fix pickle compatibility problem in different Python versions
                 Key: FLINK-22517
                 URL: https://issues.apache.org/jira/browse/FLINK-22517
             Project: Flink
          Issue Type: Bug
          Components: API / Python
    Affects Versions: 1.12.3, 1.13.0
            Reporter: Huang Xingbo
            Assignee: Huang Xingbo


Since release-1.12, PyFlink has supported Python3 8. Starting from Python 3.8, the default protocol version used by pickle is pickle5(https://www.python.org/dev/peps/pep-0574/), which will raising the following exception if the client uses python 3.8 to compile program and the cluster node uses python 3.7 or python 3.6 to run python udf:

{code:python}
ValueError: unsupported pickle protocol: 5
{code}

The workaround is to first let the python version used by the client be 3.6 or 3.7. For how to specify the client-side python execution environment, please refer to the doc(https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/python/python_config.html#python-client-executable).




--
This message was sent by Atlassian Jira
(v8.3.4#803005)