fix: Add timeout configuration for MCP servers

charts/model-engine/templates/service_template_config_map.yaml

@@ -938,6 +938,7 @@ data:
                 host: "${RESOURCE_NAME}.${NAMESPACE}.svc.cluster.local"
                 port:
                   number: 80
+          ${MCP_TIMEOUT}
   {{- end }}
   {{- if .Values.destinationrule.enabled }}
   destination-rule.yaml: |-

model-engine/model_engine_server/infra/gateways/resources/k8s_endpoint_resource_delegate.py

@@ -1,5 +1,6 @@
 import datetime
 import os
+import re
 from string import Template
 from typing import Any, Dict, List, Optional, Tuple
 

model-engine/model_engine_server/infra/gateways/resources/k8s_resource_types.py

@@ -382,6 +382,7 @@ class VirtualServiceArguments(_BaseEndpointArguments):
     """Keyword-arguments for substituting into virtual-service templates."""
 
     DNS_HOST_DOMAIN: str
+    MCP_TIMEOUT: str  # "" (Default) is 30 seconds
 
 
 class LwsServiceEntryArguments(_BaseEndpointArguments):
@@ -1361,6 +1362,25 @@ def get_endpoint_resource_arguments_from_request(
             SERVICE_NAME_OVERRIDE=service_name_override,
         )
     elif endpoint_resource_name == "virtual-service":
+        # Set 5-minute timeout for MCP servers to fix 30-second default timeout issue
+        # MCP servers use passthrough forwarder and have routes containing /mcp
+        is_mcp_server = False
+        if isinstance(flavor, RunnableImageLike) and flavor.forwarder_type == "passthrough":
+            all_routes = []
+            if flavor.predict_route:
+                all_routes.append(flavor.predict_route)
+            if flavor.routes:
+                all_routes.extend(flavor.routes)
+            if flavor.extra_routes:
+                all_routes.extend(flavor.extra_routes)
+            is_mcp_server = any("/mcp" in route.lower() for route in all_routes)
+        # Format timeout with proper indentation (10 spaces to match YAML structure)
+        # When not MCP server, use empty string
+        if is_mcp_server:
+            timeout = "          timeout: 300s"
+        else:
+            timeout = ""
+
         return VirtualServiceArguments(
             # Base resource arguments
             RESOURCE_NAME=k8s_resource_group_name,
@@ -1373,6 +1393,7 @@ def get_endpoint_resource_arguments_from_request(
             OWNER=owner,
             GIT_TAG=GIT_TAG,
             DNS_HOST_DOMAIN=infra_config().dns_host_domain,
+            MCP_TIMEOUT=timeout,
         )
     elif endpoint_resource_name == "destination-rule":
         return DestinationRuleArguments(

model-engine/model_engine_server/infra/services/live_model_endpoint_service.py

@@ -33,6 +33,9 @@
 from model_engine_server.domain.services import ModelEndpointService
 from model_engine_server.domain.use_cases.model_endpoint_use_cases import MODEL_BUNDLE_CHANGED_KEY
 from model_engine_server.infra.gateways import ModelEndpointInfraGateway
+from model_engine_server.infra.gateways.resources.k8s_endpoint_resource_delegate import (
+    K8SEndpointResourceDelegate,
+)
 from model_engine_server.infra.repositories import ModelEndpointCacheRepository
 from model_engine_server.infra.repositories.model_endpoint_record_repository import (
     ModelEndpointRecordRepository,