Good Morning. I have a GRPC server that I want to serve on Google Kubernetes Engine. My cluster already has the nginx-ingress controller installed, and I'm currently using this to serve http/https traffic. This is the ingress resource I've tried to make to host the GRPC server:
apiVersion: networking.k8s.io/v1beta1
kind: Ingress
metadata:
name: grpc-ingress-nginx
annotations:
kubernetes.io/ingress.class: "nginx"
nginx.ingress.kubernetes.io/backend-protocol: "GRPC"
cert-manager.io/cluster-issuer: "letsencrypt-prod"
nginx.ingress.kubernetes.io/grpc-backend: "true"
nginx.ingress.kubernetes.io/ssl-redirect: "true"
namespace: default
spec:
tls:
- hosts:
- bar.foo.com
secretName: reploy-tls
rules:
- host: bar.foo.com
http:
paths:
- backend:
serviceName: bar-backend-service
servicePort: 50051And here's the service/deployment for the app:
apiVersion: v1
kind: Service
metadata:
name: bar-backend-service
namespace: default
spec:
selector:
app: bar-backend-app
ports:
- port: 50051
targetPort: 50051
name: grpc
---
apiVersion: apps/v1beta1
kind: Deployment
metadata:
name: bar-backend
labels:
app: bar-backend-app
namespace: default
spec:
replicas: 1
template:
metadata:
labels:
app: bar-backend-app
spec:
containers:
- name: bar-backend-image
image: gcr.io/himank-jay/bar-backend-image:TAG
ports:
- containerPort: 50051
name: grpcWhen I run grpc_cli ls bar.foo.com:443 (using grpc_cli), I get the following error:
{"created":"@1580833741.460274000","description":"Error received from peer ipv4:x.x.x.x:x","file":"src/core/lib/surface/call.cc","file_line":1055,"grpc_message":"Socket closed","grpc_status":14}And the error from the nginx-controller is as follows:
x.x.x.x - - [04/Feb/2020:16:28:46 +0000] "PRI * HTTP/2.0" 400 157 "-" "-" 0 0.020 [] [] - - - - xxxxxAny ideas on what's wrong here? Or any thoughts on how to debug this?
The server is serving HTTP/1.x, not HTTP/2 that's required for gRPC.
You can try adding the following annotation to the Ingress config
nginx.ingress.kubernetes.io/backend-protocol: "GRPC"...as explained here.
It is also worth checking the use-http2 flag in the nginx configuration (it should be enabled, true, by default).
EDIT, regarding the new error:
PRI * HTTP/2.0 is the so-called HTTP/2 Connection Preface - part of negotiating HTTP/2. It still appears that nginx isn't configured for HTTP/2.