Enable client side gRPC health check by default

Signed-off-by: Benjamin Wang <[email protected]>

Author: ahrtr

client/v3/internal/resolver/resolver.go

@@ -15,6 +15,7 @@
 package resolver
 
 import (
+	_ "google.golang.org/grpc/health"
 	"google.golang.org/grpc/resolver"
 	"google.golang.org/grpc/resolver/manual"
 	"google.golang.org/grpc/serviceconfig"
@@ -41,7 +42,7 @@ func New(endpoints ...string) *EtcdManualResolver {
 
 // Build returns itself for Resolver, because it's both a builder and a resolver.
 func (r *EtcdManualResolver) Build(target resolver.Target, cc resolver.ClientConn, opts resolver.BuildOptions) (resolver.Resolver, error) {
-	r.serviceConfig = cc.ParseServiceConfig(`{"loadBalancingPolicy": "round_robin"}`)
+	r.serviceConfig = cc.ParseServiceConfig(`{"loadBalancingPolicy": "round_robin", "healthCheckConfig": {"serviceName": ""}}`)
 	if r.serviceConfig.Err != nil {
 		return nil, r.serviceConfig.Err
 	}

server/etcdserver/api/v3rpc/interceptor.go

@@ -34,8 +34,9 @@ import (
 )
 
 const (
-	maxNoLeaderCnt = 3
-	snapshotMethod = "/etcdserverpb.Maintenance/Snapshot"
+	maxNoLeaderCnt  = 3
+	snapshotMethod  = "/etcdserverpb.Maintenance/Snapshot"
+	gRPCHealthWatch = "/grpc.health.v1.Health/Watch"
 )
 
 type streamsMap struct {
@@ -218,7 +219,7 @@ func newStreamInterceptor(s *etcdserver.EtcdServer) grpc.StreamServerInterceptor
 			return rpctypes.ErrGRPCNotCapable
 		}
 
-		if s.IsMemberExist(s.MemberID()) && s.IsLearner() && info.FullMethod != snapshotMethod { // learner does not support stream RPC except Snapshot
+		if s.IsMemberExist(s.MemberID()) && s.IsLearner() && !isRPCStreamSupportForLearner(info) {
 			return rpctypes.ErrGRPCNotSupportedForLearner
 		}
 
@@ -259,6 +260,11 @@ func newStreamInterceptor(s *etcdserver.EtcdServer) grpc.StreamServerInterceptor
 	}
 }
 
+// learner does not support stream RPC except Snapshot and gRPC Health Watch
+func isRPCStreamSupportForLearner(info *grpc.StreamServerInfo) bool {
+	return info.FullMethod == snapshotMethod || info.FullMethod == gRPCHealthWatch
+}
+
 // cancellableContext wraps a context with new cancellable context that allows a
 // specific cancellation error to be preserved and later retrieved using the
 // Context.Err() function. This is so downstream context users can disambiguate

tests/e2e/failover_test.go

@@ -23,7 +23,6 @@ import (
 
 	"github.com/stretchr/testify/require"
 	"golang.org/x/sync/errgroup"
-	"google.golang.org/grpc"
 	_ "google.golang.org/grpc/health"
 
 	clientv3 "go.etcd.io/etcd/client/v3"
@@ -44,9 +43,8 @@ const (
 
 func TestFailoverOnDefrag(t *testing.T) {
 	tcs := []struct {
-		name            string
-		clusterOptions  []e2e.EPClusterOption
-		gRPCDialOptions []grpc.DialOption
+		name           string
+		clusterOptions []e2e.EPClusterOption
 
 		// common assertion
 		expectedMinQPS float64
@@ -62,10 +60,6 @@ func TestFailoverOnDefrag(t *testing.T) {
 				e2e.WithExperimentalStopGRPCServiceOnDefrag(true),
 				e2e.WithGoFailEnabled(true),
 			},
-			gRPCDialOptions: []grpc.DialOption{
-				grpc.WithDisableServiceConfig(),
-				grpc.WithDefaultServiceConfig(`{"loadBalancingPolicy": "round_robin", "healthCheckConfig": {"serviceName": ""}}`),
-			},
 			expectedMinQPS:         20,
 			expectedMaxFailureRate: 0.01,
 		},
@@ -76,20 +70,6 @@ func TestFailoverOnDefrag(t *testing.T) {
 				e2e.WithExperimentalStopGRPCServiceOnDefrag(false),
 				e2e.WithGoFailEnabled(true),
 			},
-			gRPCDialOptions: []grpc.DialOption{
-				grpc.WithDisableServiceConfig(),
-				grpc.WithDefaultServiceConfig(`{"loadBalancingPolicy": "round_robin", "healthCheckConfig": {"serviceName": ""}}`),
-			},
-			expectedMinQPS:         20,
-			expectedMinFailureRate: 0.25,
-		},
-		{
-			name: "defrag blocks one-third of requests with stopGRPCServiceOnDefrag set to true and client health check disabled",
-			clusterOptions: []e2e.EPClusterOption{
-				e2e.WithClusterSize(3),
-				e2e.WithExperimentalStopGRPCServiceOnDefrag(true),
-				e2e.WithGoFailEnabled(true),
-			},
 			expectedMinQPS:         20,
 			expectedMinFailureRate: 0.25,
 		},
@@ -100,10 +80,6 @@ func TestFailoverOnDefrag(t *testing.T) {
 				e2e.WithServerFeatureGate("StopGRPCServiceOnDefrag", true),
 				e2e.WithGoFailEnabled(true),
 			},
-			gRPCDialOptions: []grpc.DialOption{
-				grpc.WithDisableServiceConfig(),
-				grpc.WithDefaultServiceConfig(`{"loadBalancingPolicy": "round_robin", "healthCheckConfig": {"serviceName": ""}}`),
-			},
 			expectedMinQPS:         20,
 			expectedMaxFailureRate: 0.01,
 		},
@@ -114,20 +90,6 @@ func TestFailoverOnDefrag(t *testing.T) {
 				e2e.WithServerFeatureGate("StopGRPCServiceOnDefrag", false),
 				e2e.WithGoFailEnabled(true),
 			},
-			gRPCDialOptions: []grpc.DialOption{
-				grpc.WithDisableServiceConfig(),
-				grpc.WithDefaultServiceConfig(`{"loadBalancingPolicy": "round_robin", "healthCheckConfig": {"serviceName": ""}}`),
-			},
-			expectedMinQPS:         20,
-			expectedMinFailureRate: 0.25,
-		},
-		{
-			name: "defrag blocks one-third of requests with StopGRPCServiceOnDefrag feature gate set to true and client health check disabled",
-			clusterOptions: []e2e.EPClusterOption{
-				e2e.WithClusterSize(3),
-				e2e.WithServerFeatureGate("StopGRPCServiceOnDefrag", true),
-				e2e.WithGoFailEnabled(true),
-			},
 			expectedMinQPS:         20,
 			expectedMinFailureRate: 0.25,
 		},
@@ -151,7 +113,6 @@ func TestFailoverOnDefrag(t *testing.T) {
 					DialKeepAliveTime:    keepaliveTime,
 					DialKeepAliveTimeout: keepaliveTimeout,
 					Endpoints:            endpoints,
-					DialOptions:          tc.gRPCDialOptions,
 				})
 				if cerr != nil {
 					return cerr