fix: fixes to command stream

agent/internal/grpc/client.go

@@ -11,6 +11,7 @@ import (
 	"google.golang.org/grpc/credentials"
 	"google.golang.org/grpc/credentials/insecure"
 	"google.golang.org/grpc/encoding"
+	"google.golang.org/grpc/keepalive"
 )
 
 func init() {
@@ -26,7 +27,15 @@ func New(serverURL string, useTLS bool) (*Client, error) {
 	serverURL = strings.TrimPrefix(serverURL, "https://")
 	serverURL = strings.TrimPrefix(serverURL, "http://")
 
-	var dialOpts []grpc.DialOption
+	// Send a ping every 30s so proxies with a 60s idle timeout don't kill the
+	// long-lived CommandStream when no commands are flowing.
+	dialOpts := []grpc.DialOption{
+		grpc.WithKeepaliveParams(keepalive.ClientParameters{
+			Time:                30 * time.Second,
+			Timeout:             10 * time.Second,
+			PermitWithoutStream: false,
+		}),
+	}
 
 	if useTLS {
 		tlsCfg := &tls.Config{

server/internal/grpc/server.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"log"
 	"net"
+	"time"
 
 	"github.com/mrhid6/vantage/server/internal/grpc/pb"
 	"github.com/mrhid6/vantage/server/internal/models"
@@ -12,6 +13,7 @@ import (
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/codes"
 	"google.golang.org/grpc/encoding"
+	"google.golang.org/grpc/keepalive"
 	"google.golang.org/grpc/status"
 )
 
@@ -147,7 +149,20 @@ func StartGRPC(port int) error {
 		return fmt.Errorf("failed to listen: %w", err)
 	}
 
-	s := grpc.NewServer()
+	s := grpc.NewServer(
+		// Accept client keepalive pings as fast as every 20s so the 30s agent
+		// ping interval is always within the allowed window.
+		grpc.KeepaliveEnforcementPolicy(keepalive.EnforcementPolicy{
+			MinTime:             20 * time.Second,
+			PermitWithoutStream: false,
+		}),
+		grpc.KeepaliveParams(keepalive.ServerParameters{
+			// Server also pings the client after 45s of inactivity so both
+			// sides can detect a dead connection without waiting for a timeout.
+			Time:    45 * time.Second,
+			Timeout: 10 * time.Second,
+		}),
+	)
 	pb.RegisterVantageServer(s, &vantageServer{})
 
 	log.Printf("gRPC server listening on :%d", port)