[
https://issues.apache.org/jira/browse/BEAM-2591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16361730#comment-16361730
]
Kenneth Knowles commented on BEAM-2591:
---------------------------------------
That's about right, but maybe "implement such a client VM" is almost a noop.
Quoting [~robertwb]'s emailed reply: "I could imagine an "UniversalJavaClient"
that is parameterized by the actual runner itself dynamically. But it may also
make sense for Flink, Spark, etc. managers to serve this API directly when part
of a service."
I would start with an easy hardcode and generalize when it looks like it makes
sense. I actually think that the ReferenceRunner work may already have created
the parameterizable JobService server ([~tgroh]?)
> Python shim for submitting to FlinkRunner
> -----------------------------------------
>
> Key: BEAM-2591
> URL: https://issues.apache.org/jira/browse/BEAM-2591
> Project: Beam
> Issue Type: Sub-task
> Components: runner-flink, sdk-py-core
> Reporter: Kenneth Knowles
> Priority: Major
> Labels: portability
>
> Whatever the result of https://s.apache.org/beam-job-api, Python users will
> need to be able to pass --runner=FlinkRunner and have it work.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
