gRPC request through bash

[beatrizgoa]

Hi,

I'd like to do a gRPC request to triton server using the bash and jmeter, but I am having issues.

I have started my triton serves as:

docker run --shm-size=1g --rm -p 8000:8000 -p 8001:8001 -p 8002:8002 -v /home/models_rep:/models nvcr.io/nvidia/tritonserver:23.06-py3 tritonserver --model-repository=/models --repository-poll-secs=10 --model-control-mode=poll

and I am trying to do the request by bash like:

grpcurl -d @ -plaintext localhost:8001 inference.GRPCInferenceService.ModelInfer < load_grcp.json

but i am having the following error:

Error invoking method "inference.GRPCInferenceService.ModelInfer": failed to query for service descriptor "inference.GRPCInferenceService": server does not support the reflection API

I think that the method I am trying to use does not exist. I have been reading and maybe I need to generate a proto file, I have tried to do that from here like:

protoc --descriptor_set_out=grpc_service_descriptor.pb --include_imports common/protobuf/grpc_servi
ce.proto

And I have obtained:

model_config.proto: File not found.
common/protobuf/grpc_service.proto:33:1: Import "model_config.proto" was not found or had errors.
common/protobuf/grpc_service.proto:873:3: "ModelConfig" is not defined.

And I dont know how should I proceed to generate the bash request to be able then do it from jmeter

Thanks a lot for your help :)

[jhabaa]

Hi,
Did you found a solution ?