BatchSpanProcessor Cause a memory leak

[tensorchen]

Description

BatchSpanProcessor Cause a memory leak

Environment

• OS: [MacOS]
• Go Version: [e.g. 1.16]
• opentelemetry-go version: [v0.20.0]
• opentelemetry-go-contrib-otelgrpc [v0.20.0]

Steps To Reproduce

1. Change opentelemetry-go-contrib-otelgrpc config.Init() to the code：

driver := otlpgrpc.NewDriver(otlpgrpc.WithEndpoint("127.0.0.1:12520"))
	exporter, err := otlp.NewExporter(context.Background(), driver)
	if err != nil {
		log.Fatal(err)
	}
	kvs := []attribute.KeyValue{
		semconv.TelemetrySDKLanguageGo,
		semconv.TelemetrySDKNameKey.String("xxx"),
	}
	res := resource.NewWithAttributes(kvs...)
	tp := sdktrace.NewTracerProvider(
		sdktrace.WithSampler(sdktrace.AlwaysSample()),
		sdktrace.WithBatcher(exporter),
		sdktrace.WithResource(res),
	)
	otel.SetTracerProvider(tp)
	otel.SetTextMapPropagator(propagation.NewCompositeTextMapPropagator(propagation.TraceContext{}, propagation.Baggage{}))

2. start grpc server

3. start grpc client

ghz --insecure --proto hello-service.proto --call api.HelloService.SayHello -d '{"greeting":"tensorchen"}' -z 20m 127.0.0.1:7777

4. Result

[tianyaqu]

Duplicate with #1833 and resoleved by #1860. The cause of the leak is batch span processor will never drop the bathced spans when get an exporting failure, thus making the size of batched spans larger than the grpc message body. The batched spans will never get the chance to be successfully exported and its size grows endlessly.