Apache Beam Go SDK with Dataflow

Question

I've been working with the Go Beam SDK (v2.13.0) and can't get the wordcount example working on GCP Dataflow. It enters crash loop trying to start the org.apache.beam.runners.dataflow.worker.DataflowRunnerHarness . The example is executing correctly when run locally using the Direct runner.

The example was completely unmodified from the original example given above.

The stack trace is:

org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.InvalidProtocolBufferException: Protocol message had invalid UTF-8. 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.InvalidProtocolBufferException.invalidUtf8(InvalidProtocolBufferException.java:148) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.CodedInputStream$StreamDecoder.readStringRequireUtf8(CodedInputStream.java:2353) 
at org.apache.beam.model.pipeline.v1.RunnerApi$FunctionSpec.<init>(RunnerApi.java:59611) 
at org.apache.beam.model.pipeline.v1.RunnerApi$FunctionSpec.<init>(RunnerApi.java:59572) 
at org.apache.beam.model.pipeline.v1.RunnerApi$FunctionSpec$1.parsePartialFrom(RunnerApi.java:60241) 
at org.apache.beam.model.pipeline.v1.RunnerApi$FunctionSpec$1.parsePartialFrom(RunnerApi.java:60235) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.CodedInputStream$StreamDecoder.readMessage(CodedInputStream.java:2424) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Coder.<init>(RunnerApi.java:27531) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Coder.<init>(RunnerApi.java:27489) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Coder$1.parsePartialFrom(RunnerApi.java:28410) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Coder$1.parsePartialFrom(RunnerApi.java:28404) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Coder$Builder.mergeFrom(RunnerApi.java:28028) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Coder$Builder.mergeFrom(RunnerApi.java:27868) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.CodedInputStream$StreamDecoder.readMessage(CodedInputStream.java:2408) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.MapEntryLite.parseField(MapEntryLite.java:128) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.MapEntryLite.parseEntry(MapEntryLite.java:184) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.MapEntry.<init>(MapEntry.java:106) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.MapEntry.<init>(MapEntry.java:50) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.MapEntry$Metadata$1.parsePartialFrom(MapEntry.java:70) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.MapEntry$Metadata$1.parsePartialFrom(MapEntry.java:64) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.CodedInputStream$StreamDecoder.readMessage(CodedInputStream.java:2424) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Components.<init>(RunnerApi.java:930) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Components.<init>(RunnerApi.java:848) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Components$1.parsePartialFrom(RunnerApi.java:2714) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Components$1.parsePartialFrom(RunnerApi.java:2708) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.CodedInputStream$StreamDecoder.readMessage(CodedInputStream.java:2424) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Pipeline.<init>(RunnerApi.java:2892) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Pipeline.<init>(RunnerApi.java:2850) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Pipeline$1.parsePartialFrom(RunnerApi.java:3981) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Pipeline$1.parsePartialFrom(RunnerApi.java:3975) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.AbstractParser.parsePartialFrom(AbstractParser.java:221) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.AbstractParser.parseFrom(AbstractParser.java:239) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.AbstractParser.parseFrom(AbstractParser.java:244) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.AbstractParser.parseFrom(AbstractParser.java:49) 
at org.apache.beam.vendor.grpc.v1p13p1.com.google.protobuf.GeneratedMessageV3.parseWithIOException(GeneratedMessageV3.java:311) 
at org.apache.beam.model.pipeline.v1.RunnerApi$Pipeline.parseFrom(RunnerApi.java:3222) 
at org.apache.beam.runners.dataflow.worker.DataflowWorkerHarnessHelper.getPipelineFromEnv(DataflowWorkerHarnessHelper.java:131) 
at org.apache.beam.runners.dataflow.worker.DataflowRunnerHarness.main(DataflowRunnerHarness.java:59) 

I was using the docker image specified in the example and also tried from my own docker using the same tag (v2.13.0) but still get the same error. I Realize it's not production ready, but I am hoping the samples should work.

As per the instructions on the getting starting I ran the job like this:

wordcount --input gs://dataflow-samples/shakespeare/kinglear.txt \
--output gs://example-bucket/counts \
--runner dataflow \
--project example-project \
--temp_location gs://example-bucket/tmp/ \
--staging_location gs://example-bucket/binaries/ \
--worker_harness_container_image=apache-docker-beam-snapshots-docker.bintray.io/beam/go:20180515

Again I tried that docker provided in the getting started, as well as one built using v2.13.0.

My go.mod for the sample file is:

module example.org/wordcount

go 1.12

require (
    cloud.google.com/go v0.41.0 // indirect
    github.com/apache/beam v2.13.0+incompatible
    github.com/pkg/errors v0.8.1 // indirect
    golang.org/x/net v0.0.0-20190628185345-da137c7871d7 // indirect
    google.golang.org/grpc v1.22.0 // indirect
)

What could be causing this?

Answer 1

Dataflow doesn't officially support the Apache Beam Go SDK. Some users have had been able to use it though. I suspect that this release may have had issues. You may be able to try a different version.

You may be able to discuss with other users on the Beam mailing list about which versions have worked for them (though, unsupported).

Answer 2

I had the same problem. Increasing the Beam version to 2.19 worked for me.

Looks like they rearranged some of the Beam Fn API protos in a backwards-incompatible way so you need to make sure you're on the newest version for the Beam Go SDK to work.