Crash the server correctly during error (#2231)

test/srt/test_torch_compile.py

@@ -4,7 +4,7 @@ from types import SimpleNamespace
 
 import requests
 
-from sglang.srt.utils import kill_child_process
+from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
     DEFAULT_MODEL_NAME_FOR_TEST,
@@ -28,7 +28,7 @@ class TestTorchCompile(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid, include_self=True)
+        kill_process_tree(cls.process.pid)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_torch_compile_moe.py

@@ -4,7 +4,7 @@ from types import SimpleNamespace
 
 import requests
 
-from sglang.srt.utils import kill_child_process
+from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
     DEFAULT_SMALL_MOE_MODEL_NAME_FOR_TEST,
@@ -28,7 +28,7 @@ class TestTorchCompile(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid, include_self=True)
+        kill_process_tree(cls.process.pid)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_torchao.py

@@ -3,7 +3,7 @@ from types import SimpleNamespace
 
 import requests
 
-from sglang.srt.utils import kill_child_process
+from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
     DEFAULT_MODEL_NAME_FOR_TEST,
@@ -27,7 +27,7 @@ class TestTorchAO(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid, include_self=True)
+        kill_process_tree(cls.process.pid)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_triton_attention_backend.py

@@ -6,7 +6,7 @@ python3 -m unittest test_triton_attention_backend.TestTritonAttnBackend.test_mml
 import unittest
 from types import SimpleNamespace
 
-from sglang.srt.utils import kill_child_process
+from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
     DEFAULT_MODEL_NAME_FOR_TEST,
@@ -54,7 +54,7 @@ class TestTritonAttnBackend(unittest.TestCase):
             metrics = run_eval(args)
             self.assertGreaterEqual(metrics["score"], 0.65)
         finally:
-            kill_child_process(process.pid, include_self=True)
+            kill_process_tree(process.pid)
 
 
 if __name__ == "__main__":

test/srt/test_update_weights.py

@@ -3,7 +3,7 @@ import unittest
 
 import requests
 
-from sglang.srt.utils import kill_child_process
+from sglang.srt.utils import kill_process_tree
 from sglang.test.test_utils import (
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -23,7 +23,7 @@ class TestUpdateWeights(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid, include_self=True)
+        kill_process_tree(cls.process.pid)
 
     def run_decode(self):
         response = requests.post(

test/srt/test_vision_openai_server.py

@@ -17,7 +17,7 @@ import requests
 from decord import VideoReader, cpu
 from PIL import Image
 
-from sglang.srt.utils import kill_child_process
+from sglang.srt.utils import kill_process_tree
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
@@ -46,7 +46,7 @@ class TestOpenAIVisionServer(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid, include_self=True)
+        kill_process_tree(cls.process.pid)
 
     def test_chat_completion(self):
         client = openai.Client(api_key=self.api_key, base_url=self.base_url)
@@ -387,7 +387,7 @@ class TestQWen2VLServerContextLengthIssue(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid, include_self=True)
+        kill_process_tree(cls.process.pid)
 
     def test_chat_completion(self):
         client = openai.Client(api_key=self.api_key, base_url=self.base_url)