feat: add gRPC healthchecks (#863)

* adds a gRPC healthcheck to the "metrics" port * supports both gRPC and HTTP on "metrics" port * adds relevant doc Signed-off-by: Todd Baert <todd.baert@dynatrace.com> Co-authored-by: Michael Beemer <beeme1mr@users.noreply.github.com>
open-feature · Aug 29, 2023 · da30b7b · da30b7b
1 parent 50df791
commit da30b7b
Show file tree

Hide file tree

Showing 4 changed files with 85 additions and 41 deletions.
diff --git a/core/pkg/service/flag-evaluation/connect_service.go b/core/pkg/service/flag-evaluation/connect_service.go
@@ -7,6 +7,7 @@ import (
 	"fmt"
 	"net"
 	"net/http"
+	"strings"
 	"sync"
 	"time"
 
@@ -21,7 +22,12 @@ import (
 	"github.com/open-feature/flagd/core/pkg/telemetry"
 	"github.com/prometheus/client_golang/prometheus/promhttp"
 	"go.uber.org/zap"
+	"golang.org/x/net/http2"
+	"golang.org/x/net/http2/h2c"
 	"golang.org/x/sync/errgroup"
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/health"
+	"google.golang.org/grpc/health/grpc_health_v1"
 	"google.golang.org/protobuf/encoding/protojson"
 )
 
@@ -127,18 +133,16 @@ func (s *ConnectService) setupServer(svcConf service.Configuration) (net.Listene
 		protojson.UnmarshalOptions{DiscardUnknown: true},
 	)
 
-	path, handler := schemaConnectV1.NewServiceHandler(fes, append(svcConf.Options, marshalOpts)...)
-	mux.Handle(path, handler)
+	mux.Handle(schemaConnectV1.NewServiceHandler(fes, append(svcConf.Options, marshalOpts)...))
 
 	s.serverMtx.Lock()
 	s.server = &http.Server{
 		ReadHeaderTimeout: time.Second,
-		Handler:           handler,
+		Handler:           mux,
 	}
 	s.serverMtx.Unlock()
 
 	// Add middlewares
-
 	metricsMiddleware := metricsmw.NewHTTPMetric(metricsmw.Config{
 		Service:        svcConf.ServiceName,
 		MetricRecorder: s.metrics,
@@ -197,28 +201,41 @@ func (s *ConnectService) startServer(svcConf service.Configuration) error {
 
 func (s *ConnectService) startMetricsServer(svcConf service.Configuration) error {
 	s.logger.Info(fmt.Sprintf("metrics and probes listening at %d", svcConf.MetricsPort))
+
+	grpc := grpc.NewServer()
+	grpc_health_v1.RegisterHealthServer(grpc, health.NewServer())
+
+	mux := http.NewServeMux()
+	mux.Handle("/healthz", http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusOK)
+	}))
+	mux.Handle("/readyz", http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if s.readinessEnabled && svcConf.ReadinessProbe() {
+			w.WriteHeader(http.StatusOK)
+		} else {
+			w.WriteHeader(http.StatusPreconditionFailed)
+		}
+	}))
+	mux.Handle("/metrics", promhttp.Handler())
+
+	handler := http.HandlerFunc(func(writer http.ResponseWriter, request *http.Request) {
+		// if this is 'application/grpc' and HTTP2, handle with gRPC, otherwise HTTP.
+		if request.ProtoMajor == 2 && strings.HasPrefix(request.Header.Get("Content-Type"), "application/grpc") {
+			grpc.ServeHTTP(writer, request)
+		} else {
+			mux.ServeHTTP(writer, request)
+			return
+		}
+	})
+
 	s.metricsServerMtx.Lock()
 	s.metricsServer = &http.Server{
 		Addr:              fmt.Sprintf(":%d", svcConf.MetricsPort),
 		ReadHeaderTimeout: 3 * time.Second,
+		Handler:           h2c.NewHandler(handler, &http2.Server{}), // we need to use h2c to support plaintext HTTP2
 	}
 	s.metricsServerMtx.Unlock()
-	s.metricsServer.Handler = http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		switch r.URL.Path {
-		case "/healthz":
-			w.WriteHeader(http.StatusOK)
-		case "/readyz":
-			if s.readinessEnabled && svcConf.ReadinessProbe() {
-				w.WriteHeader(http.StatusOK)
-			} else {
-				w.WriteHeader(http.StatusPreconditionFailed)
-			}
-		case "/metrics":
-			promhttp.Handler().ServeHTTP(w, r)
-		default:
-			w.WriteHeader(http.StatusNotFound)
-		}
-	})
+
 	if err := s.metricsServer.ListenAndServe(); err != nil && !errors.Is(err, http.ErrServerClosed) {
 		return fmt.Errorf("error returned from metrics server: %w", err)
 	}

diff --git a/core/pkg/service/sync/server.go b/core/pkg/service/sync/server.go
@@ -6,15 +6,20 @@ import (
 	"fmt"
 	"net"
 	"net/http"
+	"strings"
 	"time"
 
 	rpc "buf.build/gen/go/open-feature/flagd/grpc/go/sync/v1/syncv1grpc"
 	"github.com/open-feature/flagd/core/pkg/logger"
 	iservice "github.com/open-feature/flagd/core/pkg/service"
 	syncStore "github.com/open-feature/flagd/core/pkg/sync-store"
 	"github.com/prometheus/client_golang/prometheus/promhttp"
+	"golang.org/x/net/http2"
+	"golang.org/x/net/http2/h2c"
 	"golang.org/x/sync/errgroup"
 	"google.golang.org/grpc"
+	"google.golang.org/grpc/health"
+	"google.golang.org/grpc/health/grpc_health_v1"
 )
 
 type Server struct {
@@ -101,26 +106,38 @@ func (s *Server) startServer() error {
 
 func (s *Server) startMetricsServer() error {
 	s.Logger.Info(fmt.Sprintf("binding metrics to %d", s.config.MetricsPort))
-	s.metricsServer = &http.Server{
-		ReadHeaderTimeout: 3 * time.Second,
-		Addr:              fmt.Sprintf(":%d", s.config.MetricsPort),
-	}
-	s.metricsServer.Handler = http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		switch r.URL.Path {
-		case "/healthz":
+
+	grpc := grpc.NewServer()
+	grpc_health_v1.RegisterHealthServer(grpc, health.NewServer())
+
+	mux := http.NewServeMux()
+	mux.Handle("/healthz", http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusOK)
+	}))
+	mux.Handle("/readyz", http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if s.metricServerReady && s.config.ReadinessProbe() {
 			w.WriteHeader(http.StatusOK)
-		case "/readyz":
-			if s.metricServerReady && s.config.ReadinessProbe() {
-				w.WriteHeader(http.StatusOK)
-			} else {
-				w.WriteHeader(http.StatusPreconditionFailed)
-			}
-		case "/metrics":
-			promhttp.Handler().ServeHTTP(w, r)
-		default:
-			w.WriteHeader(http.StatusNotFound)
+		} else {
+			w.WriteHeader(http.StatusPreconditionFailed)
+		}
+	}))
+	mux.Handle("/metrics", promhttp.Handler())
+
+	handler := http.HandlerFunc(func(writer http.ResponseWriter, request *http.Request) {
+		// if this is 'application/grpc' and HTTP2, handle with gRPC, otherwise HTTP.
+		if request.ProtoMajor == 2 && strings.HasPrefix(request.Header.Get("Content-Type"), "application/grpc") {
+			grpc.ServeHTTP(writer, request)
+		} else {
+			mux.ServeHTTP(writer, request)
+			return
 		}
 	})
+
+	s.metricsServer = &http.Server{
+		Addr:              fmt.Sprintf(":%d", s.config.MetricsPort),
+		ReadHeaderTimeout: 3 * time.Second,
+		Handler:           h2c.NewHandler(handler, &http2.Server{}), // we need to use h2c to support plaintext HTTP2
+	}
 	if err := s.metricsServer.ListenAndServe(); err != nil && !errors.Is(err, http.ErrServerClosed) {
 		return fmt.Errorf("error returned from metrics server: %w", err)
 	}

diff --git a/docs/other_resources/high_level_architecture.md b/docs/other_resources/high_level_architecture.md
@@ -35,14 +35,19 @@ process gets pushed to event subscribers.
 
 ## Readiness & Liveness probes
 
+### HTTP
+
 Flagd exposes HTTP liveness and readiness probes.
 These probes can be used for K8s deployments.
-With default
-start-up configurations, these probes are exposed at the following URLs,
+With default start-up configurations, these probes are exposed on the metrics port (default: 8014) at the following URLs,
 
 - Liveness: <http://localhost:8014/healthz>
 - Readiness: <http://localhost:8014/readyz>
 
+### gRPC
+
+Flagd exposes a [standard gRPC liveness check](https://github.com/grpc/grpc/blob/master/doc/health-checking.md) on the metrics port (default: 8014).
+
 ### Definition of Liveness
 
 The liveness probe becomes active and HTTP 200 status is served as soon as Flagd service is up and running.

diff --git a/web-docs/concepts/architecture.md b/web-docs/concepts/architecture.md
@@ -33,14 +33,19 @@ process gets pushed to event subscribers.
 
 ## Readiness & Liveness probes
 
+### HTTP
+
 Flagd exposes HTTP liveness and readiness probes.
 These probes can be used for K8s deployments.
-With default
-start-up configurations, these probes are exposed at the following URLs,
+With default start-up configurations, these probes are exposed on the metrics port (default: 8014) at the following URLs,
 
 - Liveness: <http://localhost:8014/healthz>
 - Readiness: <http://localhost:8014/readyz>
 
+### gRPC
+
+Flagd exposes a [standard gRPC liveness check](https://github.com/grpc/grpc/blob/master/doc/health-checking.md) on the metrics port (default: 8014).
+
 ### Definition of Liveness
 
 The liveness probe becomes active and HTTP 200 status is served as soon as Flagd service is up and running.