minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[CI] feat: add early exit to wait_for_server when process dies (#18602)

Browse Source
This commit is contained in:
shuwenn
2026-02-13 16:46:09 -08:00
committed by GitHub
parent eccf875d49
commit 3299c4f9c1
33 changed files with 229 additions and 179 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/manual/hicache/test_disaggregation_hicache.py
+2 -2
View File
@@ -33,8 +33,8 @@ class DisaggregationHiCacheBase(PDDisaggregationServerBase):
cls.start_decode()
# Block until both
cls.wait_server_ready(cls.prefill_url + "/health")
cls.wait_server_ready(cls.decode_url + "/health")
cls.wait_server_ready(cls.prefill_url + "/health", process=cls.process_prefill)
cls.wait_server_ready(cls.decode_url + "/health", process=cls.process_decode)
cls.launch_lb()
test/manual/piecewise_cudagraph/test_disaggregation_piecewise_cuda_graph.py
+2 -2
View File
@@ -25,8 +25,8 @@ class TestDisaggregationPiecewiseCudaGraph(PDDisaggregationServerBase):
cls.start_decode()
# Wait for both to be ready
cls.wait_server_ready(cls.prefill_url + "/health")
cls.wait_server_ready(cls.decode_url + "/health")
cls.wait_server_ready(cls.prefill_url + "/health", process=cls.process_prefill)
cls.wait_server_ready(cls.decode_url + "/health", process=cls.process_decode)
cls.launch_lb()
test/manual/test_mori_transfer_engine_e2e.py
+18 -6
View File
@@ -52,10 +52,14 @@ class TestMoriTransferEngineE2E(PDDisaggregationServerBase):
cls.start_decode()
cls.wait_server_ready(
cls.prefill_url + "/health", timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
cls.prefill_url + "/health",
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
process=cls.process_prefill,
)
cls.wait_server_ready(
cls.decode_url + "/health", timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
cls.decode_url + "/health",
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
process=cls.process_decode,
)
cls.launch_lb()
@@ -85,7 +89,9 @@ class TestMoriTransferEngineE2E(PDDisaggregationServerBase):
print("Starting load balancer:", " ".join(lb_command))
cls.process_lb = subprocess.Popen(lb_command, stdout=None, stderr=None)
cls.wait_server_ready(
cls.lb_url + "/health", timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
cls.lb_url + "/health",
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
process=cls.process_lb,
)
@classmethod
@@ -190,10 +196,14 @@ class TestMoriTransferEngineTPMismatchE2E(PDDisaggregationServerBase):
cls.start_decode()
cls.wait_server_ready(
cls.prefill_url + "/health", timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
cls.prefill_url + "/health",
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
process=cls.process_prefill,
)
cls.wait_server_ready(
cls.decode_url + "/health", timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
cls.decode_url + "/health",
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
process=cls.process_decode,
)
cls.launch_lb()
@@ -222,7 +232,9 @@ class TestMoriTransferEngineTPMismatchE2E(PDDisaggregationServerBase):
print("Starting load balancer:", " ".join(lb_command))
cls.process_lb = subprocess.Popen(lb_command, stdout=None, stderr=None)
cls.wait_server_ready(
cls.lb_url + "/health", timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
cls.lb_url + "/health",
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
process=cls.process_lb,
)
@classmethod