Add a watch dog thread (#1816)

python/sglang/bench_latency.py

@@ -550,4 +550,4 @@ if __name__ == "__main__":
     except Exception as e:
         raise e
     finally:
-        kill_child_process(os.getpid(), including_parent=False)
+        kill_child_process()

python/sglang/bench_server_latency.py

@@ -15,7 +15,6 @@ import dataclasses
 import itertools
 import json
 import multiprocessing
-import os
 import time
 from typing import Tuple
 
@@ -70,7 +69,7 @@ def launch_server_internal(server_args):
     except Exception as e:
         raise e
     finally:
-        kill_child_process(os.getpid(), including_parent=False)
+        kill_child_process()
 
 
 def launch_server_process(server_args: ServerArgs):
@@ -176,7 +175,7 @@ def run_benchmark(server_args: ServerArgs, bench_args: BenchArgs):
             )
     finally:
         if proc:
-            kill_child_process(proc.pid)
+            kill_child_process(proc.pid, include_self=True)
 
     print(f"\nResults are saved to {bench_args.result_filename}")
 

python/sglang/launch_server.py

@@ -15,4 +15,4 @@ if __name__ == "__main__":
     except Exception as e:
         raise e
     finally:
-        kill_child_process(os.getpid(), including_parent=False)
+        kill_child_process()

python/sglang/srt/managers/scheduler.py

@@ -18,6 +18,7 @@ limitations under the License.
 import json
 import logging
 import os
+import threading
 import time
 import warnings
 from collections import deque
@@ -222,10 +223,11 @@ class Scheduler:
         self.waiting_queue: List[Req] = []
         self.running_batch: Optional[ScheduleBatch] = None
         self.cur_batch: Optional[ScheduleBatch] = None
-        self.decode_forward_ct = 0
+        self.forward_ct = 0
-        self.stream_interval = server_args.stream_interval
+        self.forward_ct_decode = 0
         self.num_generated_tokens = 0
         self.last_stats_tic = time.time()
+        self.stream_interval = server_args.stream_interval
 
         # Init chunked prefill
         self.chunked_prefill_size = server_args.chunked_prefill_size
@@ -272,6 +274,11 @@ class Scheduler:
 
         self.batch_is_full = False
 
+        # Init watchdog thread
+        self.watchdog_timeout = server_args.watchdog_timeout
+        t = threading.Thread(target=self.watchdog_thread, daemon=True)
+        t.start()
+
         # Init profiler
         if os.getenv("SGLANG_TORCH_PROFILER_DIR", "") == "":
             self.profiler = None
@@ -289,6 +296,23 @@ class Scheduler:
                 with_stack=True,
             )
 
+    def watchdog_thread(self):
+        self.watchdog_last_forward_ct = 0
+        self.watchdog_last_time = time.time()
+
+        while True:
+            if self.cur_batch is not None:
+                if self.watchdog_last_forward_ct == self.forward_ct:
+                    if time.time() > self.watchdog_last_time + self.watchdog_timeout:
+                        logger.error(f"Watchdog timeout ({self.watchdog_timeout=})")
+                        break
+                else:
+                    self.watchdog_last_forward_ct = self.forward_ct
+                    self.watchdog_last_time = time.time()
+            time.sleep(self.watchdog_timeout / 2)
+
+        kill_parent_process()
+
     @torch.inference_mode()
     def event_loop_normal(self):
         """A normal blocking scheduler loop."""
@@ -299,6 +323,7 @@ class Scheduler:
             self.process_input_requests(recv_reqs)
 
             batch = self.get_next_batch_to_run()
+            self.cur_batch = batch
 
             if batch:
                 result = self.run_batch(batch)
@@ -746,6 +771,8 @@ class Scheduler:
 
     def run_batch(self, batch: ScheduleBatch):
         """Run a batch."""
+        self.forward_ct += 1
+
         if self.is_generation:
             if batch.forward_mode.is_decode() or batch.extend_num_tokens != 0:
                 model_worker_batch = batch.get_model_worker_batch()
@@ -778,6 +805,7 @@ class Scheduler:
             self.process_batch_result_prefill(batch, result)
 
     def process_batch_result_prefill(self, batch: ScheduleBatch, result):
+
         if self.is_generation:
             logits_output, next_token_ids, bid = result
 
@@ -890,8 +918,8 @@ class Scheduler:
 
         self.token_to_kv_pool.free_group_end()
 
-        self.decode_forward_ct = (self.decode_forward_ct + 1) % (1 << 30)
+        self.forward_ct_decode = (self.forward_ct_decode + 1) % (1 << 30)
-        if self.tp_rank == 0 and self.decode_forward_ct % 40 == 0:
+        if self.tp_rank == 0 and self.forward_ct_decode % 40 == 0:
             self.print_decode_stats()
 
     def add_logprob_return_values(
@@ -984,7 +1012,7 @@ class Scheduler:
         else:  # embedding or reward model
             output_embeddings = []
 
-        is_stream_iter = self.decode_forward_ct % self.stream_interval == 0
+        is_stream_iter = self.forward_ct_decode % self.stream_interval == 0
 
         for req in reqs:
             if req.finished() or (

python/sglang/srt/server.py

@@ -441,7 +441,7 @@ def launch_server(
 
     # Send a warmup request
     t = threading.Thread(
-        target=_wait_and_warmup, args=(server_args, pipe_finish_writer, os.getpid())
+        target=_wait_and_warmup, args=(server_args, pipe_finish_writer)
     )
     t.start()
 
@@ -496,7 +496,7 @@ def _set_envs_and_config(server_args: ServerArgs):
     mp.set_start_method("spawn", force=True)
 
 
-def _wait_and_warmup(server_args, pipe_finish_writer, pid):
+def _wait_and_warmup(server_args, pipe_finish_writer):
     headers = {}
     url = server_args.url()
     if server_args.api_key:
@@ -519,7 +519,7 @@ def _wait_and_warmup(server_args, pipe_finish_writer, pid):
         if pipe_finish_writer is not None:
             pipe_finish_writer.send(last_traceback)
         logger.error(f"Initialization failed. warmup error: {last_traceback}")
-        kill_child_process(pid, including_parent=False)
+        kill_child_process(include_self=True)
         return
 
     model_info = res.json()
@@ -551,7 +551,7 @@ def _wait_and_warmup(server_args, pipe_finish_writer, pid):
         if pipe_finish_writer is not None:
             pipe_finish_writer.send(last_traceback)
         logger.error(f"Initialization failed. warmup error: {last_traceback}")
-        kill_child_process(pid, including_parent=False)
+        kill_child_process(include_self=True)
         return
 
     # logger.info(f"{res.json()=}")
@@ -617,7 +617,7 @@ class Runtime:
 
     def shutdown(self):
         if self.pid is not None:
-            kill_child_process(self.pid)
+            kill_child_process(self.pid, include_self=True)
             self.pid = None
 
     def cache_prefix(self, prefix: str):
@@ -834,7 +834,7 @@ class Engine:
             return ret
 
     def shutdown(self):
-        kill_child_process(os.getpid(), including_parent=False)
+        kill_child_process(include_self=True)
 
     def get_tokenizer(self):
         global tokenizer_manager

python/sglang/srt/server_args.py

@@ -74,6 +74,7 @@ class ServerArgs:
     api_key: Optional[str] = None
     file_storage_pth: str = "SGLang_storage"
     enable_cache_report: bool = False
+    watchdog_timeout: float = 600
 
     # Data parallelism
     dp_size: int = 1
@@ -429,6 +430,12 @@ class ServerArgs:
             action="store_true",
             help="Return number of cached tokens in usage.prompt_tokens_details for each openai request.",
         )
+        parser.add_argument(
+            "--watchdog-timeout",
+            type=float,
+            default=ServerArgs.watchdog_timeout,
+            help="Set watchdog timeout in seconds. If a forward batch takes longer than this, the server will crash to prevent hanging.",
+        )
 
         # Data parallelism
         parser.add_argument(

python/sglang/srt/utils.py

@@ -398,17 +398,26 @@ def kill_parent_process():
     """Kill the parent process and all children of the parent process."""
     current_process = psutil.Process()
     parent_process = current_process.parent()
-    kill_child_process(parent_process.pid, skip_pid=current_process.pid)
+    kill_child_process(
-
+        parent_process.pid, include_self=True, skip_pid=current_process.pid
-
+    )
-def kill_child_process(pid, including_parent=True, skip_pid=None):
-    """Kill the process and all its children process."""
     try:
-        parent = psutil.Process(pid)
+        current_process.kill()
+    except psutil.NoSuchProcess:
+        pass
+
+
+def kill_child_process(pid=None, include_self=False, skip_pid=None):
+    """Kill the process and all its children process."""
+    if pid is None:
+        pid = os.getpid()
+
+    try:
+        itself = psutil.Process(pid)
     except psutil.NoSuchProcess:
         return
 
-    children = parent.children(recursive=True)
+    children = itself.children(recursive=True)
     for child in children:
         if child.pid == skip_pid:
             continue
@@ -417,9 +426,9 @@ def kill_child_process(pid, including_parent=True, skip_pid=None):
         except psutil.NoSuchProcess:
             pass
 
-    if including_parent:
+    if include_self:
         try:
-            parent.kill()
+            itself.kill()
         except psutil.NoSuchProcess:
             pass
 

python/sglang/test/test_utils.py

@@ -495,7 +495,7 @@ def run_unittest_files(files: List[str], timeout_per_file: float):
             )
             assert ret_code == 0
         except TimeoutError:
-            kill_child_process(process.pid)
+            kill_child_process(process.pid, include_self=True)
             time.sleep(5)
             print(
                 f"\nTimeout after {timeout_per_file} seconds when running {filename}\n",
@@ -563,7 +563,7 @@ def run_bench_serving(
     try:
         res = run_benchmark(args)
     finally:
-        kill_child_process(process.pid)
+        kill_child_process(process.pid, include_self=True)
 
     assert res["completed"] == num_prompts
     return res
@@ -596,7 +596,7 @@ def run_bench_latency(model, other_args):
         lastline = output.split("\n")[-3]
         output_throughput = float(lastline.split(" ")[-2])
     finally:
-        kill_child_process(process.pid)
+        kill_child_process(process.pid, include_self=True)
 
     return output_throughput
 
@@ -707,8 +707,8 @@ def run_mmlu_test(
         pass
 
     # Clean up everything
-    kill_child_process(process.pid)
+    kill_child_process(process.pid, include_self=True)
-    kill_child_process(process.pid)
+    kill_child_process(process.pid, include_self=True)
     stdout.close()
     stderr.close()
     if os.path.exists(STDOUT_FILENAME):

test/srt/sampling/penaltylib/test_srt_endpoint_with_penalizers.py

@@ -31,7 +31,7 @@ class TestBatchPenalizerE2E(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def run_decode(
         self,

test/srt/test_cache_report.py

@@ -45,7 +45,7 @@ class TestCacheReport(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def run_decode(self, return_logprob=False, top_logprobs_num=0, n=1):
         response = requests.post(

test/srt/test_data_parallelism.py

@@ -25,7 +25,7 @@ class TestDataParallelism(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_double_sparsity.py

@@ -43,7 +43,7 @@ class TestDoubleSparsity(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_embedding_openai_server.py

@@ -28,7 +28,7 @@ class TestOpenAIServer(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def run_embedding(self, use_list_input, token_input):
         client = openai.Client(api_key=self.api_key, base_url=self.base_url)

test/srt/test_eval_accuracy_large.py

@@ -30,7 +30,7 @@ class TestEvalAccuracyLarge(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_eval_accuracy_large_chunked_prefill.py

@@ -25,7 +25,7 @@ class TestEvalAccuracyLargeChunkedPrefill(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_eval_accuracy_large_mixed_chunked_prefill.py

@@ -31,7 +31,7 @@ class TestEvalAccuracyLargeChunkedPrefill(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_eval_accuracy_mini.py

@@ -22,7 +22,7 @@ class TestEvalAccuracyMini(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_json_constrained.py

@@ -41,7 +41,7 @@ class TestJSONConstrained(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def run_decode(self, json_schema, return_logprob=False, top_logprobs_num=0, n=1):
         response = requests.post(

test/srt/test_large_max_new_tokens.py

@@ -42,7 +42,7 @@ class TestLargeMaxNewTokens(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
         cls.stdout.close()
         cls.stderr.close()
         os.remove("stdout.txt")

test/srt/test_matched_stop.py

@@ -32,7 +32,7 @@ class TestMatchedStop(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def run_completions_generation(
         self,

test/srt/test_mla.py

@@ -25,7 +25,7 @@ class TestMLA(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_mla_fp8.py

@@ -31,7 +31,7 @@ class TestMLA(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mgsm_en(self):
         args = SimpleNamespace(

test/srt/test_moe_eval_accuracy_large.py

@@ -35,7 +35,7 @@ class TestMoEEvalAccuracyLarge(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_nightly_gsm8k_eval.py

@@ -36,7 +36,7 @@ class TestEvalAccuracyLarge(unittest.TestCase):
 
     def tearDown(self):
         if self.process:
-            kill_child_process(self.process.pid)
+            kill_child_process(self.process.pid, include_self=True)
 
     def launch_server(self, model, is_fp8, is_tp2):
         other_args = ["--log-level-http", "warning", "--trust-remote-code"]

test/srt/test_openai_server.py

@@ -31,7 +31,7 @@ class TestOpenAIServer(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def run_completion(
         self, echo, logprobs, use_list_input, parallel_sample_num, token_input

test/srt/test_pytorch_sampling_backend.py

@@ -27,7 +27,7 @@ class TestPyTorchSamplingBackend(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_retract_decode.py

@@ -22,7 +22,7 @@ class TestRetractDecode(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_skip_tokenizer_init.py

@@ -26,7 +26,7 @@ class TestSkipTokenizerInit(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def run_decode(self, return_logprob=False, top_logprobs_num=0, n=1):
         max_new_tokens = 32

test/srt/test_srt_endpoint.py

@@ -27,7 +27,7 @@ class TestSRTEndpoint(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def run_decode(
         self,

test/srt/test_torch_compile.py

@@ -27,7 +27,7 @@ class TestTorchCompile(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_torchao.py

@@ -27,7 +27,7 @@ class TestTorchCompile(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_mmlu(self):
         args = SimpleNamespace(

test/srt/test_triton_attn_backend.py

@@ -50,7 +50,7 @@ class TestTritonAttnBackend(unittest.TestCase):
             metrics = run_eval(args)
             assert metrics["score"] >= 0.65
         finally:
-            kill_child_process(process.pid)
+            kill_child_process(process.pid, include_self=True)
 
 
 if __name__ == "__main__":

test/srt/test_update_weights.py

@@ -23,7 +23,7 @@ class TestUpdateWeights(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def run_decode(self):
         response = requests.post(

test/srt/test_vision_openai_server.py

@@ -45,7 +45,7 @@ class TestOpenAIVisionServer(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        kill_child_process(cls.process.pid)
+        kill_child_process(cls.process.pid, include_self=True)
 
     def test_chat_completion(self):
         client = openai.Client(api_key=self.api_key, base_url=self.base_url)