fix: makes gRPC's MaxRecvMsgSize for the extension server (envoyproxy#1291)

**Description**
Makes go's gRPC `MaxRecvMsgSize` configurable for the extension server,
whose default value is 4MB
https://pkg.go.dev/google.golang.org/grpc#MaxRecvMsgSize

When the gRPC message payload from the xDS server exceeds 4MB, it is
rejected by the extension server and the below errors is observed on
envoy-gateway:
```
2025-10-07T10:26:44.625Z        ERROR   xds     runner/runner.go:245    failed to translate xds ir      {"runner": "xds", "error": "rpc error: code = ResourceExhausted desc = grpc: received message larger than max (6683527 vs. 4194304)"}
2025-10-07T10:26:44.625Z        ERROR   watchable       message/watchutil.go:86 observed an error       {"runner": "xds", "error": "rpc error: code = ResourceExhausted desc = grpc: received message larger than max (6683527 vs. 4194304)"}
```
This PR makes it configurable via the command-line flag `maxRecvMsgSize`
and keeps its default at 4MB.

**Related Issues/PRs (if applicable)**
Fixes envoyproxy#1293
Related PR to make gRPC's message sizes configurable, though that was
for the external processor gRPC server:
envoyproxy#1213

**Special notes for reviewers (if applicable)**

Note that this applies to the xDS resource messages sent from the xDS
server to the extension server. The payload of these messages increases
with the size/count of the config resources.

---------

Signed-off-by: Owayss Kabtoul <[email protected]>
Signed-off-by: Hrushikesh Patil <[email protected]>

Author: owayss

cmd/controller/main.go

@@ -49,6 +49,8 @@ type flags struct {
 	extProcExtraEnvVars            string
 	// extProcMaxRecvMsgSize is the maximum message size in bytes that the gRPC server can receive.
 	extProcMaxRecvMsgSize int
+	// maxRecvMsgSize is the maximum message size in bytes that the gRPC extension server can receive.
+	maxRecvMsgSize int
 }
 
 // parsePullPolicy parses string into a k8s PullPolicy.
@@ -146,6 +148,11 @@ func parseAndValidateFlags(args []string) (flags, error) {
 		512*1024*1024,
 		"Maximum message size in bytes that the gRPC server can receive for extProc. Default is 512MB.",
 	)
+	maxRecvMsgSize := fs.Int(
+		"maxRecvMsgSize",
+		4*1024*1024,
+		"Maximum message size in bytes that the gRPC extension server can receive. Default is 4MB.",
+	)
 
 	if err := fs.Parse(args); err != nil {
 		err = fmt.Errorf("failed to parse flags: %w", err)
@@ -215,6 +222,7 @@ func parseAndValidateFlags(args []string) (flags, error) {
 		rootPrefix:                     *rootPrefix,
 		extProcExtraEnvVars:            *extProcExtraEnvVars,
 		extProcMaxRecvMsgSize:          *extProcMaxRecvMsgSize,
+		maxRecvMsgSize:                 *maxRecvMsgSize,
 	}, nil
 }
 
@@ -271,7 +279,7 @@ func main() {
 
 	// Start the extension server running alongside the controller.
 	const extProcUDSPath = "/etc/ai-gateway-extproc-uds/run.sock"
-	s := grpc.NewServer()
+	s := grpc.NewServer(grpc.MaxRecvMsgSize(flags.maxRecvMsgSize))
 	extSrv := extensionserver.New(mgr.GetClient(), ctrl.Log, extProcUDSPath, false)
 	egextension.RegisterEnvoyGatewayExtensionServer(s, extSrv)
 	grpc_health_v1.RegisterHealthServer(s, extSrv)

cmd/controller/main_test.go

@@ -29,6 +29,7 @@ func Test_parseAndValidateFlags(t *testing.T) {
 		require.Equal(t, "/certs", f.tlsCertDir)
 		require.Equal(t, "tls.crt", f.tlsCertName)
 		require.Equal(t, "tls.key", f.tlsKeyName)
+		require.Equal(t, 4*1024*1024, f.maxRecvMsgSize)
 		require.NoError(t, err)
 	})
 	t.Run("all flags", func(t *testing.T) {
@@ -49,6 +50,7 @@ func Test_parseAndValidateFlags(t *testing.T) {
 					tc.dash + "port=:8080",
 					tc.dash + "extProcExtraEnvVars=OTEL_SERVICE_NAME=test;OTEL_TRACES_EXPORTER=console",
 					tc.dash + "spanRequestHeaderAttributes=x-session-id:session.id",
+					tc.dash + "maxRecvMsgSize=33554432",
 				}
 				f, err := parseAndValidateFlags(args)
 				require.Equal(t, "debug", f.extProcLogLevel)
@@ -59,6 +61,7 @@ func Test_parseAndValidateFlags(t *testing.T) {
 				require.Equal(t, ":8080", f.extensionServerPort)
 				require.Equal(t, "OTEL_SERVICE_NAME=test;OTEL_TRACES_EXPORTER=console", f.extProcExtraEnvVars)
 				require.Equal(t, "x-session-id:session.id", f.spanRequestHeaderAttributes)
+				require.Equal(t, 32*1024*1024, f.maxRecvMsgSize)
 				require.NoError(t, err)
 			})
 		}

manifests/charts/ai-gateway-helm/templates/deployment.yaml

@@ -62,6 +62,9 @@ spec:
             - --enableLeaderElection=true
             {{- end }}
             - --rootPrefix={{ .Values.endpointConfig.rootPrefix }}
+            {{- if ne .Values.controller.maxRecvMsgSize nil }}
+            - --maxRecvMsgSize={{ .Values.controller.maxRecvMsgSize }}
+            {{- end }}
           livenessProbe:
             grpc:
               port: 1063

manifests/charts/ai-gateway-helm/values.yaml

@@ -155,4 +155,8 @@ controller:
   nodeSelector: {}
   tolerations: []
   affinity: {}
-
+  # maxRecvMsgSize is the maximum message size in bytes that the gRPC extension server can receive
+  # from xDS (envoy-gateway).
+  # This value should be increased in setups where count/complexity of xDS
+  # resources (configuration) is big. Defaults to 4MB.
+  # maxRecvMsgSize: "4194304"