[Config] Retire get_global_server_args, and clear the deprecated flags that have a replacement (#38375)

python/sglang/test/ascend/test_npu_logging.py

@@ -25,7 +25,7 @@ class TestNPULoggingBase(CustomTestCase):
     [Test Target] --log-requests; --log-requests-level; --log-requests-target; --uvicorn-access-log-exclude-prefixes;
     --enable-metrics; --enable-metrics-for-all-scheduler;
     --bucket-time-to-first-token; --bucket-inter-token-latency; --bucket-e2e-request-latency;
-    --collect-tokens-histogram; --prompt-tokens-buckets; --generation-tokens-buckets;
+    --prompt-tokens-buckets; --generation-tokens-buckets;
     --tokenizer-metrics-custom-labels-header; --tokenizer-metrics-allowed-custom-labels;
     --gc-warning-threshold-secs
     """

python/sglang/test/chunked_prefill_test_utils.py

@@ -34,7 +34,7 @@ KV_CANARY_ARGS: List[str] = [
     "partial",
     "--kv-canary-sweep-interval",
     "100",
-    "--disable-piecewise-cuda-graph",
+    "--cuda-graph-backend-prefill=disabled",
 ]
 
 

python/sglang/test/kv_canary/utils.py

@@ -17,7 +17,7 @@ def build_canary_server_args(
     args = [
         "--kv-canary",
         kv_canary_mode.value,
-        "--disable-piecewise-cuda-graph",
+        "--cuda-graph-backend-prefill=disabled",
         "--context-length",
         "16384",
         *extra_server_args,

python/sglang/test/mock_model/utils.py

@@ -21,7 +21,7 @@ _MOCK_MODEL_SERVER_ARGS_NO_CANARY: list[str] = [
     "dummy",
     "--sampling-backend",
     "token_oracle",
-    "--disable-piecewise-cuda-graph",
+    "--cuda-graph-backend-prefill=disabled",
 ]