Collector return "invalid length for ID" for trace reporting

[jacky15]

Describe the bug
I am using opentelemetry-collector with otlphttp exporter. But when data report to the collector, it return HTTP status code 400 with the message "invalid length for ID"

Steps to reproduce
Here is the tcp traffic

POST /v1/traces HTTP/1.1
Host: 10.0.1.21:4317
User-Agent: Go-http-client/1.1
Content-Length: 20316
Content-Type: application/json
Accept-Encoding: gzip

{"resourceSpans":[{"resource":{"attributes":[{"key":"service.name","value":{"stringValue":"unknown_service:unify-query"}},{"key":"telemetry.sdk.language","value":{"stringValue":"go"}},{"key":"telemetry.sdk.name","value":{"stringValue":"opentelemetry"}},{"key":"telemetry.sdk.version","value":{"stringValue":"0.19.0"}}]},"instrumentationLibrarySpans":[{"instrumentationLibrary":{"name":"bk-monitorv3/unify-query"},"spans":[{"traceId":"cgXI6uihz643E6Zqq7Nl4w==","spanId":"+iPdFTeJDX8=","name":"/metrics","kind":"SPAN_KIND_INTERNAL","startTimeUnixNano":"1621253293188761633","endTimeUnixNano":"1621253293190601010","status":{"code":"STATUS_CODE_OK"}},{"traceId":"tXrmFgvi148s6EsroJ1S1A==","spanId":"zIqhNylUXuk=","name":"/metrics","kind":"SPAN_KIND_INTERNAL","startTimeUnixNano":"1621254568188753964","endTimeUnixNano":"1621254568189747171","status":{"code":"STATUS_CODE_OK"}},{"traceId":"Q2sKzOrO09THC8n5rb8V5Q==","spanId":"x6GjCl4l8t8=","name":"/metrics","kind":"SPAN_KIND_INTERNAL","startTimeUnixNano":"1621254583188746771","endTimeUnixNano":"1621254583189715582","status":{"code":"STATUS_CODE_OK"}}]}]}]}

HTTP/1.1 400 Bad Request
Content-Type: application/json
Date: Mon, 17 May 2021 12:29:56 GMT
Content-Length: 73

{
  "code": 3,
  "message": "invalid length for ID",
  "details": [
  ]
}

What did you expect to see?
Report success.

What did you see instead?
Report failed with 400 status code

What version did you use?
Version: c1e3d47

What config did you use?

receivers:
  otlp:
    protocols:
      http:
        endpoint: "0.0.0.0:4317"
processors:
  batch:
extensions:
  health_check: {}
exporters:
  zipkin:
    endpoint: "http://10.0.0.1:9411/api/v2/spans"
    format: proto
  logging:
    loglevel: debug
extensions:
  health_check:
  pprof:
    endpoint: :1888
  zpages:
    endpoint: :55679
service:
  extensions: [pprof, zpages, health_check]
  pipelines:
    traces:
      receivers: [otlp]
      exporters: [zipkin]
      processors: [batch]

Environment
OS: centos-release-7-8.2003.0.el7.centos.x86_64
Compiler(if manually compiled): go version go1.14 linux/amd64

[bogdandrutu]

the trace/span ID need to be hex encoded not base64 cgXI6uihz643E6Zqq7Nl4w==. See https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/protocol/otlp.md#otlphttp-request

[bryanuribe]

Do you need someone to work this? If so, would I be able to pick this up?
cc: @alolita

[jacky15]

the trace/span ID need to be hex encoded not base64 cgXI6uihz643E6Zqq7Nl4w==. See https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/protocol/otlp.md#otlphttp-request

According to this, Is this a bug of the http exporter?

[bogdandrutu]

What http exporter do you use? Seems like a bug there if they send JSON data (which seems to be the case based on the payload)

[jacky15]

I am using the exporter from "go.opentelemetry.io/otel/exporters/otlp/otlphttp"
Here is the code used to init the exporter

func (s *Service) Start(ctx context.Context) {
	var err error

	if !viper.GetBool(EnableConfigPath) {
		log.SugaredLogger.Infof("trace is disable, no exporter will start.")
		return
	}

	s.driverOps = []otlphttp.Option{
		otlphttp.WithEndpoint(strings.Join([]string{reportHost, reportPort}, ":")),
		otlphttp.WithInsecure(),
		otlphttp.WithMarshal(otlphttp.MarshalJSON),
	}
	s.traceDriver = otlphttp.NewDriver(s.driverOps...)

	if s.Exporter, err = otlp.NewExporter(ctx, s.traceDriver); err != nil {
		log.SugaredLogger.Errorf("failed to config tracing exporter for err->[%s]", err)
		return
	}
	s.wg.Add(1)

	bsp := sdktrace.NewBatchSpanProcessor(s.Exporter)
	tp := sdktrace.NewTracerProvider(sdktrace.WithSpanProcessor(bsp))
	otel.SetTracerProvider(tp)
	propagator := propagation.NewCompositeTextMapPropagator(propagation.Baggage{}, propagation.TraceContext{})
	otel.SetTextMapPropagator(propagator)

	s.ctx, s.cancelFunc = context.WithCancel(ctx)
	log.SugaredLogger.Infof("trace exporter start success.")
}

[bogdandrutu]

Oh, so you are using golang sdk. Just change for the moment:

otlphttp.WithMarshal(otlphttp.MarshalJSON) to otlphttp.WithMarshal(otlphttp.MarshalProto)

[bogdandrutu]

Created a bug in the golang SDK, closing this since we implement the specs.

[jacky15]

It works after change the code to otlphttp.WithMarshal(otlphttp.MarshalProto)
Thanks a lot!

[utezduyar]

Actually I think the marshaled output is correct, base64 according to the spec: https://github.com/open-telemetry/oteps/blob/main/text/0122-otlp-http-json.md#json-mapping
Use proto3 standard defined [JSON Mapping](https://developers.google.com/protocol-buffers/docs/proto3#json) for mapping between protobuf and json. trace_id and span_id is base64 encoded in OTLP/HTTP+JSON, not hex.
There is a mismatch between the spec and the collector.