The DirectRunner error looks like it's because the FnApiRunner doesn't support SDF.
What is the coder id for the Flink error? It looks like the full stack trace should contain it. On Mon, Aug 3, 2020 at 10:09 AM Piotr Szuberski <[email protected]> wrote: > I'm Writing SpannerIO.Write cross-language transform and when I try to run > it from python I receive errors: > > On Flink: > apache_beam.utils.subprocess_server: INFO: b'Caused by: > java.lang.IllegalArgumentException: Transform external_1HolderCoder uses > unknown accumulator coder id %s' > apache_beam.utils.subprocess_server: INFO: b'\tat > org.apache.beam.vendor.guava.v26_0_jre.com.google.common.base.Preconditions.checkArgument(Preconditions.java:216)' > apache_beam.utils.subprocess_server: INFO: b'\tat > org.apache.beam.runners.core.construction.graph.PipelineValidator.validateCombine(PipelineValidator.java:273)' > > On DirectRunner: > File > "/Users/piotr/beam/sdks/python/apache_beam/runners/portability/fn_api_runner/fn_runner.py", > line 181, in run_via_runner_api > self._validate_requirements(pipeline_proto) > File > "/Users/piotr/beam/sdks/python/apache_beam/runners/portability/fn_api_runner/fn_runner.py", > line 264, in _validate_requirements > raise ValueError( > ValueError: Missing requirement declaration: > {'beam:requirement:pardo:splittable_dofn:v1'} > > I suppose that SpannerIO.Write uses a transform that cannot be translated > in cross-language usage? I'm not sure whether there is something I can do > about it. > >
