@Yang, Yes, you are right. Python jobs could be seen as special JAR jobs whose main class is always `org.apache.flink.client.python.PythonDriver`. What we could do in Flink K8s operator is to make it more convenient and handle properly for the different kinds of dependencies[1].
@Gyula, I can help on this. I will find some time to investigate this in the following days and will let you know when there is any progress. Regards, Dian [1] https://nightlies.apache.org/flink/flink-docs-master/docs/dev/python/dependency_management/ On Mon, Jul 4, 2022 at 11:52 AM Yang Wang <danrtsey...@gmail.com> wrote: > AFAIK, the python job could be considered as a special case of jar job. > The user jar is flink-python-*.jar and is located in the opt directory. > The python script is just the argument of this user jar. So I believe the > users already could submit python jobs via Flink Kubernetes operator. > However, they need some manual operations, including specify the user jar, > download python script via init container, etc. > > What we could do in the Flink kubernetes operator is to make the > submission more convenient by introducing a new field(e.g. pyScript). > > cc @Dian Fu <dian0511...@gmail.com> @biaoge...@gmail.com > <biaoge...@gmail.com> WDYT? > > Best, > Yang > > Gyula Fóra <gyf...@apache.org> 于2022年7月4日周一 00:12写道： > >> Hi Devs! >> >> Would anyone with a good understanding of the Python execution layer be >> interested in working on adding Python job support for the Flink >> Kubernetes >> Operator? >> >> This is a feature request that comes up often ( >> https://issues.apache.org/jira/browse/FLINK-28364) and it would be a >> great >> way to fill some missing feature gaps on the operator :) >> >> I am of course happy to help or work together with someone on this but I >> have zero experience with the Python API at this stage and don't want to >> miss some obvious requirements. >> >> Cheers, >> Gyula >> >
