Skip to content

Latest commit

 

History

History
142 lines (108 loc) · 4.88 KB

README.md

File metadata and controls

142 lines (108 loc) · 4.88 KB
Raw

goHelloGrpcStream

a test Hello world application using GRPC streaming in GO

Usage

Proto file

use protoc if you need to re-generate the proto files. Refer to https://github.com/grpc/grpc-go/tree/master/examples if you need help.

cd helloworld
protoc -I helloworld/ helloworld/helloworld.proto --go_out=plugins=grpc:helloworld

Server

The server opens a TCP socket and wait for GRPC messages to come in

    cd helloworld/greeter_server/
    go build
    ./greeter_server

2017/11/13 10:35:54 Listening on tcp://localhost: :7788

The server does not support HTTPS yet. You have to use a middleware (Istio/Envoy, Traefik, Nginx...) to handle the TLS termination and send plain HTTP/2 to the server

Client

The client connect to the server on the provided host:port and :

  • request a hello world message and display the return from the server
  • request a hello world streaming message and display each message until the server stops sending
cd helloworld/greeter_client/
go build
./greeter_client -debug -server=localhost:7788 -unary

2017/11/13 10:26:04 Greeting: Hello world:7788
2017/11/13 10:26:09 Hello Streamworld:7788
2017/11/13 10:26:14 Hello Streamworld:7788
2017/11/13 10:26:19 Hello Streamworld:7788

The client support two modes :

  • unary : will send a single HTTP/2 request This is usefull to test the low level connectivity
  • stream : will send a HTTP/2 di-directional Stream request and keep the stream opened This is usefull to test the longevity of the connection and the number of possible parallel connections

The client support TLS, see -h for options

Loadtest

The loadtest application opens one HTTP/2 streaming connection per -clients and maintain it for -cnxDelay

The loadtest support TLS, see -h for options

Docker

Use the docker file to build an image embedding both client and server code. Best is to use the makefile :

make docker
make docker-push

or by hand :

docker build -t prune/gohellogrpcstream:latest .
docker push prune/gohellogrpcstream:latest

Kubernetes

Deploys a server + service that listen on port 7788 along with a load-tester client

    kubectl -n dev apply -f kubernetes/deployment-autoinject-istio.yml

The client pod will not do anything so Istio have the time to settle. Start the loadtest_client using :

kubectl exec -ti $(kubectl get pods --selector=app=loadtest-client -o jsonpath='{.items..metadata.name}')  /root/loadtest_client

Load Test

client

Use the loadtest_client which is simulating any number of clients in parallel. The loadtest_client make one TCP connexion per client which then open a gRPC HTTP/2 Stream. Starting 100 clients should show you 100 POST requests on /helloworld.Greeter/SayHelloStream. When stopping the client, the 100 streams will be closed, showing 100 /helloworld.Greeter/SayHelloStream POSTs.

You can then scale the number of connexions by either :

  • editing the number of clients from the ENV
  • scaling to more pods

...


# Istio
This tool was primilary made to test Istio Service Mesh deployed in Kubernetes.
in the Kubernetes folder you will find the needed files to deploy the pods and create the Istio config.

## Using Ingress (Istio < 0.8.0)
You can test the client/server inside K8s but from an external point of view using an Ingress.
Update the `kubernetes/ingress.yml` file with a public DNS name you own then deploy using :

kubectl -n dev apply -f kubernetes/ingress.yml


Then update the `greeter-client` or `loadtest-client` to use the external endpoint :

... containers: - name: loadtest-client image: "prune/gohellogrpcstream:latest"
ports: - containerPort: 7788 name: loadtest-client command: - "/root/loadtest_client" env: - name: "CLIENTS" value: "10" - name: "SERVER" value: "greet.dev.mydomain.com:80"

You may use the Destination Policy and Destination Rule files to better configure the limits of the Istio Ingress. Note that Istio < 1.x have some issues with the Max Concurent Streams values.

Using Istio > 0.8.0 (including 1.0.0+)

In 0.8.0 and 1.x, Istio changed his setup and no longer use the Kubernetes Ingress. Instead, Istio is using a LoadBalancer Service to forward all the traffic to Istio IngressGateway. We here consider you have a working Istio Service Mesh. Refer to Istio docs otherwise.

On the Istio side, you now need to create a Gateway that will allow incoming traffic to the Mesh and a VirtualService that will forward this traffic to a Kubernetes Service.

Deploy the gateway.yml then the virtualservice.yml. You should now be able to use the greeter_client or the loadtest_client by pointing it to hello.test.domain.ca:80 or hello.test.domain.ca:443. Don't forger to replace all the demo domain names by your own, and propagate the DNS changes :)