chore: rename terminate_existing to terminate_all_matching_processes (#5923)

Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>
Co-authored-by: Keiven Chang <keivenchang@users.noreply.github.com>

tests/basic/test_autodeploy_backend.py

@@ -73,7 +73,7 @@ class DynamoWorkerProcess(ManagedProcess):
             ],
             timeout=360,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             log_dir=log_dir,
         )
 

tests/conftest.py

@@ -278,7 +278,7 @@ class EtcdServer(ManagedProcess):
             command=command,
             timeout=timeout,
             display_output=False,
-            terminate_existing=not use_random_port,  # Disabled for parallel test execution with random ports
+            terminate_all_matching_process_names=not use_random_port,  # For distributed tests, do not terminate all matching processes
             health_check_ports=[port],
             data_dir=data_dir,
             log_dir=request.node.name,
@@ -325,7 +325,7 @@ class NatsServer(ManagedProcess):
             command=command,
             timeout=timeout,
             display_output=False,
-            terminate_existing=not use_random_port,  # Disabled for parallel test execution with random ports
+            terminate_all_matching_process_names=not use_random_port,  # For distributed tests, do not terminate all matching processes
             data_dir=data_dir,
             health_check_ports=[port],
             health_check_funcs=[self._nats_ready],

tests/fault_tolerance/cancellation/test_sglang.py

@@ -142,7 +142,7 @@ class DynamoWorkerProcess(ManagedProcess):
             health_check_urls=health_check_urls,
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             # Ensure any orphaned SGLang engine cores or child helpers are cleaned up
             stragglers=[
                 "SGLANG:EngineCore",

tests/fault_tolerance/cancellation/test_trtllm.py

@@ -134,7 +134,7 @@ class DynamoWorkerProcess(ManagedProcess):
             health_check_urls=health_check_urls,
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             log_dir=log_dir,
         )
 

tests/fault_tolerance/cancellation/test_vllm.py

@@ -123,7 +123,7 @@ class DynamoWorkerProcess(ManagedProcess):
             health_check_urls=health_check_urls,
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             # Ensure any orphaned vLLM engine cores or child helpers are cleaned up
             stragglers=[
                 "VLLM::EngineCore",

tests/fault_tolerance/cancellation/utils.py

@@ -36,7 +36,7 @@ class DynamoFrontendProcess(BaseDynamoFrontendProcess):
             frontend_port=0,  # allocate a free port (xdist-safe)
             router_mode="round-robin",
             extra_env=extra_env,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
         )
 
 

tests/fault_tolerance/etcd_ha/test_sglang.py

@@ -118,7 +118,7 @@ class DynamoWorkerProcess(ManagedProcess):
             health_check_urls=health_check_urls,
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             # Ensure any orphaned SGLang engine cores or child helpers are cleaned up
             stragglers=[
                 "SGLANG:EngineCore",

tests/fault_tolerance/etcd_ha/test_trtllm.py

@@ -107,7 +107,7 @@ class DynamoWorkerProcess(ManagedProcess):
             health_check_urls=health_check_urls,
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             log_dir=log_dir,
         )
 

tests/fault_tolerance/etcd_ha/test_vllm.py

@@ -84,7 +84,7 @@ class DynamoWorkerProcess(ManagedProcess):
             health_check_urls=health_check_urls,
             timeout=120,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             stragglers=[
                 "VLLM::EngineCore",
             ],

tests/fault_tolerance/etcd_ha/utils.py

@@ -31,11 +31,16 @@ class DynamoFrontendProcess(BaseDynamoFrontendProcess):
             "DYN_LOG": "debug",
             "ETCD_ENDPOINTS": ",".join(etcd_endpoints),
         }
+        # WARNING: terminate_all_matching_process_names=True is NOT pytest-xdist safe!
+        # DANGER: Kills ALL dynamo-frontend processes system-wide, including other parallel tests.
+        # For parallel-safe alternative, use terminate_all_matching_process_names=False.
+        # See tests/kvbm_integration/common.py:llm_server_kvbm for example.
+        # TODO: Switch to terminate_all_matching_process_names=False with dynamic ports
         super().__init__(
             request,
             router_mode="round-robin",
             extra_env=extra_env,
-            terminate_existing=True,
+            terminate_all_matching_process_names=True,  # TODO: Change to False
         )
 
 
@@ -90,7 +95,7 @@ class EtcdReplicaServer(ManagedProcess):
             command=command,
             timeout=timeout,
             display_output=False,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             data_dir=data_dir,
             log_dir=log_dir,
         )

tests/fault_tolerance/gpu_memory_service/utils/common.py

@@ -96,7 +96,7 @@ class GMSServerProcess(ManagedProcess):
             env={**os.environ, "DYN_LOG": "debug"},
             timeout=60,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             log_dir=log_dir,
             health_check_funcs=[self._socket_ready],
         )

tests/fault_tolerance/gpu_memory_service/utils/sglang.py

@@ -60,7 +60,7 @@ class SGLangWithGMSProcess(ManagedProcess):
             ],
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             stragglers=[],
             log_dir=log_dir,
         )

tests/fault_tolerance/gpu_memory_service/utils/vllm.py

@@ -61,7 +61,7 @@ class VLLMWithGMSProcess(ManagedProcess):
             ],
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             stragglers=[],
             log_dir=log_dir,
         )

tests/fault_tolerance/migration/test_sglang.py

@@ -184,7 +184,7 @@ class DynamoWorkerProcess(ManagedProcess):
             health_check_urls=health_check_urls,
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             stragglers=["SGLANG:EngineCore"],
             straggler_commands=["-m dynamo.sglang"],
             log_dir=log_dir,

tests/fault_tolerance/migration/test_trtllm.py

@@ -173,7 +173,7 @@ class DynamoWorkerProcess(ManagedProcess):
             health_check_urls=health_check_urls,
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             log_dir=log_dir,
             display_name=worker_id,
         )

tests/fault_tolerance/migration/test_vllm.py

@@ -167,7 +167,7 @@ class DynamoWorkerProcess(ManagedProcess):
             health_check_urls=health_check_urls,
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             stragglers=["VLLM::EngineCore"],
             straggler_commands=["-m dynamo.vllm"],
             log_dir=log_dir,

tests/fault_tolerance/migration/utils.py

@@ -37,7 +37,7 @@ class DynamoFrontendProcess(BaseDynamoFrontendProcess):
             router_mode="round-robin",
             extra_args=extra_args if extra_args else None,
             extra_env=extra_env,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             display_name="frontend",
         )
 

tests/fault_tolerance/test_vllm_health_check.py

@@ -63,7 +63,7 @@ class DynamoWorkerProcess(ManagedProcess):
             ],
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             stragglers=["VLLM::EngineCore"],
             straggler_commands=["-m dynamo.vllm"],
             log_dir=log_dir,

tests/frontend/conftest.py

@@ -58,7 +58,7 @@ def start_services_with_http(
     with DynamoFrontendProcess(
         request,
         frontend_port=ports.frontend_port,
-        terminate_existing=False,
+        terminate_all_matching_process_names=False,
     ):
         logger.info(f"HTTP Frontend started on port {ports.frontend_port}")
         yield ports.frontend_port, ports.system_ports[0]
@@ -188,7 +188,7 @@ def start_services_with_grpc(
         with DynamoFrontendProcess(
             request,
             frontend_port=ports.frontend_port,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             extra_args=[
                 "--kserve-grpc-server",
                 "--grpc-metrics-port",
@@ -261,7 +261,7 @@ class MockerWorkerProcess(ManagedProcess):
             ],
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             stragglers=["VLLM::EngineCore"],
             straggler_commands=["-m dynamo.mocker"],
             log_dir=log_dir,

tests/frontend/grpc/test_tensor_mocker_engine.py

@@ -63,7 +63,7 @@ class MockWorkerProcess(ManagedProcess):
             ],
             timeout=300,
             display_output=True,
-            terminate_existing=False,
+            terminate_all_matching_process_names=False,
             stragglers=[],
             straggler_commands=["echo_tensor_worker.py"],
             log_dir=log_dir,