Details
-
Improvement
-
Status: Resolved
-
P2
-
Resolution: Fixed
-
None
Description
The default behavior of the Flink and Spark Python runner clients is to look for the job server jar at a path relative to the source file [1]. This will break whenever we change how the Python source is installed in our tests (such as [2]). The default behavior can be overridden by passing an absolute path to --flink_job_server_jar [3], which we should do.
[1] https://github.com/apache/beam/blob/d4818ecc56d5b1a54bfc5f05ae0fd49f4142364c/sdks/python/apache_beam/utils/subprocess_server.py#L162
[2] https://github.com/apache/beam/pull/10378
[3] https://github.com/apache/beam/blob/bb794884c8f856c8ff38ac903f60d4e0b6a05f15/sdks/python/apache_beam/options/pipeline_options.py#L986