refactor(load-snapshot): build LoadSnapshot directly, drop legacy get_loads IPC (#30525)

Co-authored-by: cctry <cctry@fb.com>

test/registered/unit/entrypoints/test_v1_loads_aggregate.py

@@ -15,7 +15,9 @@ from sglang.srt.managers.load_snapshot import (
     SLOT_LEN_STRUCT,
     SLOT_SIZE,
     VERSION,
+    DisaggregationMetrics,
     LoadSnapshot,
+    QueueMetrics,
     ShmLoadSnapshotReader,
     ShmLoadSnapshotWriter,
     slot_offset,
@@ -150,14 +152,10 @@ class TestGetLoads(CustomTestCase):
                     cache_hit_rate=0.75,
                     utilization=0.5,
                     max_running_requests=128,
-                    has_disaggregation=1,
-                    disagg_mode=2,
-                    decode_transfer_queue_reqs=4,
-                    has_queues=1,
-                    queue_waiting=2,
-                    queue_grammar=1,
-                    queue_paused=0,
-                    queue_retracted=3,
+                    disaggregation=DisaggregationMetrics(
+                        mode="decode", decode_transfer_queue_reqs=4
+                    ),
+                    queues=QueueMetrics(waiting=2, grammar=1, paused=0, retracted=3),
                 )
             )
 

test/registered/utils/test_type_based_dispatcher.py

@@ -34,7 +34,6 @@ class TestTypeBasedDispatcher(unittest.TestCase):
             FlushCacheReqInput,
             FreezeGCReq,
             GetInternalStateReq,
-            GetLoadsReqInput,
             GetWeightsByNameReqInput,
             InitWeightsSendGroupForRemoteInstanceReqInput,
             InitWeightsUpdateGroupReqInput,
@@ -114,7 +113,6 @@ class TestTypeBasedDispatcher(unittest.TestCase):
             (ExpertDistributionReq, lambda req: "expert_distribution_handled"),
             (LoadLoRAAdapterReqInput, lambda req: "load_lora_adapter_handled"),
             (UnloadLoRAAdapterReqInput, lambda req: "unload_lora_adapter_handled"),
-            (GetLoadsReqInput, lambda req: "get_loads_handled"),
         ]
 
         # Create requests that conforms to the real distribution
@@ -209,7 +207,6 @@ class TestTypeBasedDispatcher(unittest.TestCase):
         test_requests.append(GetWeightsByNameReqInput(name=""))
         test_requests.append(ReleaseMemoryOccupationReqInput())
         test_requests.append(RpcReqInput(method=""))
-        test_requests.append(GetLoadsReqInput())
 
         dispatcher = TypeBasedDispatcher(mapping)