Deploying gRPC service in Google Cloud Run

gRPC + Cloud Run

REST is the most popular interface to access a web service, gRPC is a modern high performance framework that can run in any environment. If you remember protocol buffers, gRPC is the most straightforward way to use it.

Google Cloud Run is a serverless platform which can host any container with HTTP endpoints (now gRPC too) and it takes care of scaling, authentication and monitoring on its own. Hence its language agnostic and all you need to know how to package your software in a container exposing an endpoint.

In this article we will create a simple service which exposes a gRPC endpoint in NodeJS and host it in Google Cloud Run.

First thing, we define the interface in which data is exchanged. So we create a proto file. Complete file ping.proto.

This contains not just the message format (like a JSON contract of a REST service) but also the operations that is allowed (like methods in a SOAP WSDL)

service PingServer {
   rpc doPing(Ping) returns (Pong) {}

message Ping {
    string greetings = 1;

message Pong {
    string acknowledgement = 1;

Next we create the server. We create a main function that starts the gRPC server

function main() {
  var server = new grpc.Server();
  server.addService(ping_proto.PingServer.service, {doPing: doPing});

  const port = parseInt(process.env.PORT) || 50051;

  server.bindAsync(''+port, grpc.ServerCredentials.createInsecure(), () => {
    console.log("Starting server in port "+port);

Now that we have added the service doPing, lets add an implementation.

function doPing(call, callback) {
  console.log("Request received "+call.request);
  callback(null, {acknowledgement: 'Thank you for ' + call.request.greetings});

Here is the complete server code. ping-server.js

We can run and test the server locally

Lets run the server locally

node ./pingserver/ping-server.js &

We can also build a dcoker image and run so that we know that its packaged properly and good to be deployed in cloud. This is the beauty of containers, once the image is created it would behave exacly same in any other host like it does locally, so a lot of environment setup issues can be ruled out while debugging.

docker build . -t$GCP_PROJECT/grpc-ping:latest
docker run -d -p 50051:50051 -e PORT=50051$GCP_PROJECT/grpc-ping          

To invoke the method on the runnning service we could either

grpcurl \                                                                                        
    -plaintext -proto protos/ping.proto \
    -d '{"greetings": "Hello"}' \
    localhost:50051 \

You should see a response from the server

Response: Thank you for Namaste

Now we can deploy this in one of the projects in Google Cloud Platform.

gcloud auth login ## Authentocate 
gcloud builds submit --tag$GOOGLE_CLOUD_PROJECT/grpc-ping ## build the image 
cloud auth configure-docker ## login to the container image registry
docker run -d -p 50051:50051 -e PORT=50051$GOOGLE_CLOUD_PROJECT/grpc-ping ## run the container locally
gcloud run deploy ping-upstream --image$GOOGLE_CLOUD_PROJECT/grpc-ping

Again using grpcurl we could test invoke the service

grpcurl \                                                                                        
    -proto protos/ping.proto \           
    -d '{"greetings": "Hello"}' \ \

If you are trying to call the service hosted in cloud run you would need SSL in the client code grpc.credentials.createSsl() because cloud run ingress adds a default SSL layer.

The latest version of Postman also supports gRPC, if you like a nice UI. It doesn’t work in Postman web due to browser limitation of gRPC, it needs the postman agent at least.

If you want to consume such gRPC service from a Flutter client, Debkanchan Samadder has created a minimal demo app DebkanchanSamadder/flutter-grpc-demo consuming it.

If you like the post please spread the word