[vtadmin] grpc healthserver + channelz (#10038)

* Add `--rbac` to example vtadmin-up

This was introduced and made a required flag (with mutually-exclusive `--no-rbac`)
in #9972.

Signed-off-by: Andrew Mason <andrew@planetscale.com>

* Cleanup `Options.Services`, and inspect the ServiceInfo() directly

Instead of requiring users to know in advance (and put into their startup
scripts), which services will run, we can (correctly) inspect the service
names that have been registered on the underlying grpc.Server and mark
them as SERVING in the health service.

_Alsooooo_, this was never bound to a flag, so users couldn't use it
anyway! Whoops.

Signed-off-by: Andrew Mason <andrew@planetscale.com>

* Provide option to enable channelz service

Signed-off-by: Andrew Mason <andrew@planetscale.com>

* Add flag to enable channelz

Signed-off-by: Andrew Mason <andrew@planetscale.com>

* Tidy up flags; label all sections

Signed-off-by: Andrew Mason <andrew@planetscale.com>

* Add missing flag for AllowTracing

Signed-off-by: Andrew Mason <andrew@planetscale.com>

examples/local/scripts/vtadmin-up.sh

@@ -14,6 +14,7 @@ vtadmin \
   --http-tracing \
   --logtostderr \
   --alsologtostderr \
+  --rbac \
   --rbac-config="./vtadmin/rbac.yaml" \
   --cluster "id=local,name=local,discovery=staticfile,discovery-staticfile-path=./vtadmin/discovery.json,tablet-fqdn-tmpl={{ .Tablet.Hostname }}:15{{ .Tablet.Alias.Uid }}" \
   > "${log_dir}/vtadmin-api.out" 2>&1 &

go/cmd/vtadmin/main.go

@@ -141,20 +141,29 @@ func run(cmd *cobra.Command, args []string) {
 }
 
 func main() {
+	// Common flags
 	rootCmd.Flags().StringVar(&opts.Addr, "addr", ":15000", "address to serve on")
 	rootCmd.Flags().DurationVar(&opts.CMuxReadTimeout, "lmux-read-timeout", time.Second, "how long to spend connection muxing")
 	rootCmd.Flags().DurationVar(&opts.LameDuckDuration, "lame-duck-duration", time.Second*5, "length of lame duck period at shutdown")
+
+	// Cluster config flags
 	rootCmd.Flags().Var(&clusterConfigs, "cluster", "per-cluster configuration. any values here take precedence over those in -cluster-defaults or -cluster-config")
 	rootCmd.Flags().Var(&clusterFileConfig, "cluster-config", "path to a yaml cluster configuration. see clusters.example.yaml") // (TODO:@amason) provide example config.
 	rootCmd.Flags().Var(&defaultClusterConfig, "cluster-defaults", "default options for all clusters")
 
+	// Tracing flags
 	rootCmd.Flags().AddGoFlag(flag.Lookup("tracer"))                 // defined in go/vt/trace
 	rootCmd.Flags().AddGoFlag(flag.Lookup("tracing-enable-logging")) // defined in go/vt/trace
 	rootCmd.Flags().AddGoFlag(flag.Lookup("tracing-sampling-type"))  // defined in go/vt/trace
 	rootCmd.Flags().AddGoFlag(flag.Lookup("tracing-sampling-rate"))  // defined in go/vt/trace
 	rootCmd.Flags().BoolVar(&opts.EnableTracing, "grpc-tracing", false, "whether to enable tracing on the gRPC server")
 	rootCmd.Flags().BoolVar(&httpOpts.EnableTracing, "http-tracing", false, "whether to enable tracing on the HTTP server")
 
+	// gRPC server flags
+	rootCmd.Flags().BoolVar(&opts.AllowReflection, "grpc-allow-reflection", false, "whether to register the gRPC server for reflection; this is required to use tools like `grpc_cli`")
+	rootCmd.Flags().BoolVar(&opts.EnableChannelz, "grpc-enable-channelz", false, "whether to enable the channelz service on the gRPC server")
+
+	// HTTP server flags
 	rootCmd.Flags().BoolVar(&httpOpts.DisableCompression, "http-no-compress", false, "whether to disable compression of HTTP API responses")
 	rootCmd.Flags().BoolVar(&httpOpts.DisableDebug, "http-no-debug", false, "whether to disable /debug/pprof/* and /debug/env HTTP endpoints")
 	rootCmd.Flags().Var(&debug.OmitEnv, "http-debug-omit-env", "name of an environment variable to omit from /debug/env, if http debug endpoints are enabled. specify multiple times to omit multiple env vars")
@@ -169,7 +178,7 @@ func main() {
 	)
 	rootCmd.Flags().BoolVar(&httpOpts.EnableDynamicClusters, "http-enable-dynamic-clusters", false, "whether to enable dynamic clusters that are set by request header cookies")
 
-	// rbac flags
+	// RBAC flags
 	rootCmd.Flags().StringVar(&rbacConfigPath, "rbac-config", "", "path to an RBAC config file. must be set if passing --rbac")
 	rootCmd.Flags().BoolVar(&enableRBAC, "rbac", false, "whether to enable RBAC. must be set if not passing --rbac")
 	rootCmd.Flags().BoolVar(&disableRBAC, "no-rbac", false, "whether to disable RBAC. must be set if not passing --no-rbac")

go/vt/vtadmin/api.go

@@ -106,10 +106,6 @@ func NewAPI(clusters []*cluster.Cluster, opts Options) *API {
 		return c1.ID < c2.ID
 	}).Sort(clusters)
 
-	if opts.GRPCOpts.Services == nil {
-		opts.GRPCOpts.Services = []string{"vtadmin.VTAdminServer"}
-	}
-
 	var (
 		authn rbac.Authenticator
 		authz *rbac.Authorizer

go/vt/vtadmin/grpcserver/server.go

@@ -34,6 +34,7 @@ import (
 	"github.com/opentracing/opentracing-go"
 	"github.com/soheilhy/cmux"
 	"google.golang.org/grpc"
+	channelz "google.golang.org/grpc/channelz/service"
 	"google.golang.org/grpc/health"
 	"google.golang.org/grpc/reflection"
 
@@ -63,26 +64,14 @@ type Options struct {
 	// EnableTracing specifies whether to install opentracing interceptors on
 	// the gRPC server.
 	EnableTracing bool
-	// Services is a list of service names to declare as SERVING in health
-	// checks. Names should be fully-qualified (package_name.service_name, e.g.
-	// vtadmin.VTAdminServer, not VTAdminServer), and must be unique for a
-	// single Server instance. Users of this package are responsible for
-	// ensuring they do not pass a list with duplicate service names.
-	//
-	// The service name "grpc.health.v1.Health" is reserved by this package in
-	// order to power the healthcheck service. Attempting to pass this in the
-	// Services list to a grpcserver will be ignored.
-	//
-	// See https://github.com/grpc/grpc/blob/7324556353e831c57d30973db33df489c3ed3576/doc/health-checking.md
-	// for more details on healthchecking.
-	Services []string
+	// EnableChannelz specifies whether to register the channelz service on the
+	// gRPC server.
+	EnableChannelz bool
 
 	StreamInterceptors []grpc.StreamServerInterceptor
 	UnaryInterceptors  []grpc.UnaryServerInterceptor
 }
 
-const healthServiceName = "grpc.health.v1.Health" // reserved health service name
-
 // Server provides a multiplexed gRPC/HTTP server.
 type Server struct {
 	name string
@@ -137,6 +126,10 @@ func New(name string, opts Options) *Server {
 	healthServer := health.NewServer()
 	healthpb.RegisterHealthServer(gserv, healthServer)
 
+	if opts.EnableChannelz {
+		channelz.RegisterChannelzServiceToServer(gserv)
+	}
+
 	return &Server{
 		name:         name,
 		gRPCServer:   gserv,
@@ -225,15 +218,8 @@ func (s *Server) ListenAndServe() error { // nolint:funlen
 		shutdown <- err
 	}()
 
-	s.healthServer.SetServingStatus(healthServiceName, healthpb.HealthCheckResponse_SERVING)
-
-	for _, name := range s.opts.Services {
-		if name == healthServiceName {
-			log.Warningf("Attempted to register a service under the reserved healthcheck service name %s; ignoring", healthServiceName)
-			continue
-		}
-
-		s.healthServer.SetServingStatus(name, healthpb.HealthCheckResponse_SERVING)
+	for service := range s.gRPCServer.GetServiceInfo() {
+		s.healthServer.SetServingStatus(service, healthpb.HealthCheckResponse_SERVING)
 	}
 
 	s.setServing(true)