fix: fix config updates on deep copy not propagating back (#3512)

Signed-off-by: hongkuanz <hongkuanz@nvidia.com>

benchmarks/profiler/profile_sla.py

@@ -901,4 +901,14 @@ if __name__ == "__main__":
     )
     args = parser.parse_args()
 
+    # setup file logging
+    os.makedirs(args.output_dir, exist_ok=True)
+    log_file_handler = logging.FileHandler(f"{args.output_dir}/profile_sla.log")
+    log_file_handler.setLevel(logging.INFO)
+    formatter = logging.Formatter(
+        "%(asctime)s - %(name)s - %(levelname)s - %(message)s", "%Y-%m-%d %H:%M:%S"
+    )
+    log_file_handler.setFormatter(formatter)
+    logger.addHandler(log_file_handler)
+
     asyncio.run(run_profile(args))

benchmarks/profiler/utils/config.py

@@ -224,14 +224,14 @@ def set_multinode_config(worker_service, gpu_count: int, num_gpus_per_node: int)
 
 
 def get_service_name_by_type(
-    config: dict, backend: str, sub_component_type: SubComponentType
+    config: Config, backend: str, sub_component_type: SubComponentType
 ) -> str:
     """Helper function to get service name by subComponentType.
 
     First tries to find service by subComponentType, then falls back to component name.
 
     Args:
-        config: Configuration dictionary (with spec.services structure)
+        config: Configuration object
         backend: Backend name (e.g., "sglang", "vllm", "trtllm")
         sub_component_type: The type of sub-component to look for (PREFILL or DECODE)
 
@@ -239,11 +239,7 @@ def get_service_name_by_type(
         The service name
     """
     # Check if config has the expected structure
-    if (
-        not isinstance(config, dict)
-        or "spec" not in config
-        or "services" not in config.get("spec", {})
-    ):
+    if not config.spec or not config.spec.services:
         # Fall back to default name if structure is unexpected
         if sub_component_type == SubComponentType.DECODE:
             return WORKER_COMPONENT_NAMES[backend].decode_worker_k8s_name
@@ -251,12 +247,9 @@ def get_service_name_by_type(
             return WORKER_COMPONENT_NAMES[backend].prefill_worker_k8s_name
 
     # Look through services to find one with matching subComponentType
-    services = config["spec"]["services"]
+    services = config.spec.services
     for service_name, service_config in services.items():
-        if (
-            isinstance(service_config, dict)
-            and service_config.get("subComponentType") == sub_component_type.value
-        ):
+        if service_config.subComponentType == sub_component_type.value:
             return service_name
 
     # Fall back to default component names
@@ -274,7 +267,7 @@ def get_service_name_by_type(
 
 
 def get_worker_service_from_config(
-    config: dict,
+    config: Config,
     backend: str = "sglang",
     sub_component_type: SubComponentType = SubComponentType.DECODE,
 ):
@@ -299,8 +292,7 @@ def get_worker_service_from_config(
     service_name = get_service_name_by_type(config, backend, sub_component_type)
 
     # Get the actual service from the config
-    cfg = Config.model_validate(config)
-    return cfg.spec.services[service_name]
+    return config.spec.services[service_name]
 
 
 def setup_worker_service_resources(
@@ -445,10 +437,10 @@ class VllmV1ConfigModifier:
         if target == "prefill":
             # Get service names by inferring from subComponentType first
             prefill_service_name = get_service_name_by_type(
-                config, "vllm", SubComponentType.PREFILL
+                cfg, "vllm", SubComponentType.PREFILL
             )
             decode_service_name = get_service_name_by_type(
-                config, "vllm", SubComponentType.DECODE
+                cfg, "vllm", SubComponentType.DECODE
             )
 
             # convert prefill worker into decode worker
@@ -461,7 +453,7 @@ class VllmV1ConfigModifier:
             cfg.spec.services[decode_service_name].subComponentType = "decode"
 
             worker_service = get_worker_service_from_config(
-                cfg.model_dump(),
+                cfg,
                 backend="vllm",
                 sub_component_type=SubComponentType.DECODE,
             )
@@ -482,10 +474,10 @@ class VllmV1ConfigModifier:
         elif target == "decode":
             # Get service names by inferring from subComponentType first
             prefill_service_name = get_service_name_by_type(
-                config, "vllm", SubComponentType.PREFILL
+                cfg, "vllm", SubComponentType.PREFILL
             )
             decode_service_name = get_service_name_by_type(
-                config, "vllm", SubComponentType.DECODE
+                cfg, "vllm", SubComponentType.DECODE
             )
 
             # delete prefill worker
@@ -495,7 +487,7 @@ class VllmV1ConfigModifier:
             cfg.spec.services[decode_service_name].subComponentType = "decode"
 
             worker_service = get_worker_service_from_config(
-                cfg.model_dump(),
+                cfg,
                 backend="vllm",
                 sub_component_type=SubComponentType.DECODE,
             )
@@ -513,7 +505,7 @@ class VllmV1ConfigModifier:
         # set num workers to 1
         # Use the inferred decode service name
         final_decode_service_name = get_service_name_by_type(
-            cfg.model_dump(), "vllm", SubComponentType.DECODE
+            cfg, "vllm", SubComponentType.DECODE
         )
         decode_worker_config = cfg.spec.services[final_decode_service_name]
         decode_worker_config.replicas = 1
@@ -529,7 +521,7 @@ class VllmV1ConfigModifier:
     ):
         cfg = Config.model_validate(config)
         worker_service = get_worker_service_from_config(
-            config, backend="vllm", sub_component_type=component_type
+            cfg, backend="vllm", sub_component_type=component_type
         )
 
         # Set up resources
@@ -575,8 +567,9 @@ class VllmV1ConfigModifier:
 
     @classmethod
     def get_model_name(cls, config: dict) -> str:
+        cfg = Config.model_validate(config)
         try:
-            worker_service = get_worker_service_from_config(config, backend="vllm")
+            worker_service = get_worker_service_from_config(cfg, backend="vllm")
             args = validate_and_get_worker_args(worker_service, backend="vllm")
         except (ValueError, KeyError):
             logger.warning(
@@ -670,10 +663,10 @@ class SGLangConfigModifier:
         if target == "prefill":
             # Get service names by inferring from subComponentType first
             prefill_service_name = get_service_name_by_type(
-                config, "sglang", SubComponentType.PREFILL
+                cfg, "sglang", SubComponentType.PREFILL
             )
             decode_service_name = get_service_name_by_type(
-                config, "sglang", SubComponentType.DECODE
+                cfg, "sglang", SubComponentType.DECODE
             )
 
             # convert prefill worker into decode worker
@@ -686,7 +679,7 @@ class SGLangConfigModifier:
             cfg.spec.services[decode_service_name].subComponentType = "decode"
 
             worker_service = get_worker_service_from_config(
-                cfg.model_dump(),
+                cfg,
                 backend="sglang",
                 sub_component_type=SubComponentType.DECODE,
             )
@@ -707,10 +700,10 @@ class SGLangConfigModifier:
         elif target == "decode":
             # Get service names by inferring from subComponentType first
             prefill_service_name = get_service_name_by_type(
-                config, "sglang", SubComponentType.PREFILL
+                cfg, "sglang", SubComponentType.PREFILL
             )
             decode_service_name = get_service_name_by_type(
-                config, "sglang", SubComponentType.DECODE
+                cfg, "sglang", SubComponentType.DECODE
             )
 
             # delete prefill worker
@@ -720,7 +713,7 @@ class SGLangConfigModifier:
             cfg.spec.services[decode_service_name].subComponentType = "decode"
 
             worker_service = get_worker_service_from_config(
-                cfg.model_dump(),
+                cfg,
                 backend="sglang",
                 sub_component_type=SubComponentType.DECODE,
             )
@@ -751,7 +744,7 @@ class SGLangConfigModifier:
         # set num workers to 1
         # Use the inferred decode service name
         final_decode_service_name = get_service_name_by_type(
-            cfg.model_dump(), "sglang", SubComponentType.DECODE
+            cfg, "sglang", SubComponentType.DECODE
         )
         decode_worker_config = cfg.spec.services[final_decode_service_name]
         decode_worker_config.replicas = 1
@@ -767,7 +760,7 @@ class SGLangConfigModifier:
     ):
         cfg = Config.model_validate(config)
         worker_service = get_worker_service_from_config(
-            config, backend="sglang", sub_component_type=component_type
+            cfg, backend="sglang", sub_component_type=component_type
         )
 
         # Set up resources
@@ -792,7 +785,7 @@ class SGLangConfigModifier:
     ):
         cfg = Config.model_validate(config)
         worker_service = get_worker_service_from_config(
-            config, backend="sglang", sub_component_type=component_type
+            cfg, backend="sglang", sub_component_type=component_type
         )
 
         # Set up resources with multinode configuration
@@ -827,7 +820,7 @@ class SGLangConfigModifier:
     ):
         cfg = Config.model_validate(config)
         worker_service = get_worker_service_from_config(
-            config, backend="sglang", sub_component_type=component_type
+            cfg, backend="sglang", sub_component_type=component_type
         )
 
         # Set up resources with multinode configuration
@@ -854,8 +847,9 @@ class SGLangConfigModifier:
 
     @classmethod
     def get_model_name(cls, config: dict) -> str:
+        cfg = Config.model_validate(config)
         try:
-            worker_service = get_worker_service_from_config(config, backend="sglang")
+            worker_service = get_worker_service_from_config(cfg, backend="sglang")
             args = validate_and_get_worker_args(worker_service, backend="sglang")
         except (ValueError, KeyError):
             logger.warning(
@@ -946,10 +940,10 @@ class TrtllmConfigModifier:
         if target == "prefill":
             # Get service names by inferring from subComponentType first
             prefill_service_name = get_service_name_by_type(
-                config, "trtllm", SubComponentType.PREFILL
+                cfg, "trtllm", SubComponentType.PREFILL
             )
             decode_service_name = get_service_name_by_type(
-                config, "trtllm", SubComponentType.DECODE
+                cfg, "trtllm", SubComponentType.DECODE
             )
 
             # Convert to prefill-only aggregated setup
@@ -963,7 +957,7 @@ class TrtllmConfigModifier:
             cfg.spec.services[decode_service_name].subComponentType = "decode"
 
             worker_service = get_worker_service_from_config(
-                cfg.model_dump(),
+                cfg,
                 backend="trtllm",
                 sub_component_type=SubComponentType.DECODE,
             )
@@ -1000,10 +994,10 @@ class TrtllmConfigModifier:
         elif target == "decode":
             # Get service names by inferring from subComponentType first
             prefill_service_name = get_service_name_by_type(
-                config, "trtllm", SubComponentType.PREFILL
+                cfg, "trtllm", SubComponentType.PREFILL
             )
             decode_service_name = get_service_name_by_type(
-                config, "trtllm", SubComponentType.DECODE
+                cfg, "trtllm", SubComponentType.DECODE
             )
 
             # Convert to decode-only aggregated setup
@@ -1015,7 +1009,7 @@ class TrtllmConfigModifier:
 
             # Decode worker already has the correct name
             worker_service = get_worker_service_from_config(
-                cfg.model_dump(),
+                cfg,
                 backend="trtllm",
                 sub_component_type=SubComponentType.DECODE,
             )
@@ -1048,7 +1042,7 @@ class TrtllmConfigModifier:
         # Set num workers to 1
         # Use the inferred decode service name
         final_decode_service_name = get_service_name_by_type(
-            cfg.model_dump(), "trtllm", SubComponentType.DECODE
+            cfg, "trtllm", SubComponentType.DECODE
         )
         worker_config = cfg.spec.services[final_decode_service_name]
         worker_config.replicas = 1
@@ -1067,7 +1061,7 @@ class TrtllmConfigModifier:
         # Get the worker service using helper function
         # This assumes convert_config has been called, so the service is named decode_worker_k8s_name
         worker_service = get_worker_service_from_config(
-            config, backend="trtllm", sub_component_type=component_type
+            cfg, backend="trtllm", sub_component_type=component_type
         )
 
         # Set up resources
@@ -1118,8 +1112,9 @@ class TrtllmConfigModifier:
 
     @classmethod
     def get_model_name(cls, config: dict) -> str:
+        cfg = Config.model_validate(config)
         try:
-            worker_service = get_worker_service_from_config(config, backend="trtllm")
+            worker_service = get_worker_service_from_config(cfg, backend="trtllm")
             args = validate_and_get_worker_args(worker_service, backend="trtllm")
         except (ValueError, KeyError):
             logger.warning(

components/backends/sglang/deploy/disagg.yaml

@@ -18,6 +18,7 @@ spec:
       envFromSecret: hf-token-secret
       dynamoNamespace: sglang-disagg
       componentType: worker
+      subComponentType: decode
       replicas: 1
       resources:
         limits:
@@ -50,6 +51,7 @@ spec:
       envFromSecret: hf-token-secret
       dynamoNamespace: sglang-disagg
       componentType: worker
+      subComponentType: prefill
       replicas: 1
       resources:
         limits:
@@ -59,7 +61,7 @@ spec:
           image: my-registry/sglang-runtime:my-tag
           workingDir: /workspace/components/backends/sglang
           command:
-          - python3
+          - python3E
           - -m
           - dynamo.sglang
           args:

components/backends/trtllm/deploy/disagg.yaml

@@ -18,6 +18,7 @@ spec:
       dynamoNamespace: trtllm-disagg
       envFromSecret: hf-token-secret
       componentType: worker
+      subComponentType: prefill
       replicas: 1
       resources:
         limits:
@@ -45,6 +46,7 @@ spec:
       dynamoNamespace: trtllm-disagg
       envFromSecret: hf-token-secret
       componentType: worker
+      subComponentType: decode
       replicas: 1
       resources:
         limits:

components/backends/vllm/deploy/disagg.yaml

@@ -18,6 +18,7 @@ spec:
       dynamoNamespace: vllm-disagg
       envFromSecret: hf-token-secret
       componentType: worker
+      subComponentType: decode
       replicas: 1
       resources:
         limits:
@@ -37,6 +38,7 @@ spec:
       dynamoNamespace: vllm-disagg
       envFromSecret: hf-token-secret
       componentType: worker
+      subComponentType: prefill
       replicas: 1
       resources:
         limits: