[refactor] Move ServerArgs ownership into the runtime context (stack 2/15) (#30064)

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

python/sglang/multimodal_gen/test/unit/test_disagg_trace.py

@@ -33,6 +33,7 @@ from sglang.multimodal_gen.runtime.pipelines_core import Req
 from sglang.srt import server_args as srt_server_args_module
 from sglang.srt.observability import trace as srt_trace
 from sglang.srt.observability.trace import TraceNullContext, TraceReqContext
+from sglang.srt.runtime_context import reset_context
 from sglang.srt.server_args import set_global_server_args_for_scheduler
 
 try:
@@ -62,12 +63,18 @@ def _enable_minimal_otel() -> None:
 
 @contextmanager
 def _srt_trace_server_args():
-    prev_server_args = srt_server_args_module._global_server_args
+    try:
+        prev_server_args = srt_server_args_module.get_global_server_args()
+    except ValueError:  # nothing published yet
+        prev_server_args = None
     set_global_server_args_for_scheduler(SimpleNamespace(trace_modules="request"))
     try:
         yield
     finally:
-        srt_server_args_module._global_server_args = prev_server_args
+        if prev_server_args is None:
+            reset_context()
+        else:
+            set_global_server_args_for_scheduler(prev_server_args)
 
 
 def _traceparent_from(ctx) -> str | None:

python/sglang/srt/runtime_context.py

@@ -21,10 +21,12 @@ wrapper, not a cache. It gives call-sites one import and one naming scheme in
 place of a dozen free functions, plus a test-only ``override()`` hook to force a
 topology without monkeypatching the underlying getters.
 
-``get_server_args()`` returns the process-wide ``ServerArgs`` (the config tier).
-It is a read-through to ``server_args.get_global_server_args()`` — same object,
-same pre-publish error — so new code can adopt the context accessor while the
-legacy getter remains canonical.
+``get_server_args()`` returns the process-wide ``ServerArgs`` (the config
+tier). The context owns the storage: publishing goes through
+``RuntimeContext.set_server_args`` (the legacy
+``set_global_server_args_for_scheduler`` / ``get_global_server_args`` in
+``server_args.py`` are thin shims over this slot), and the object is returned
+by reference — the same live instance everywhere, never a copy.
 """
 
 from __future__ import annotations
@@ -50,12 +52,6 @@ def _dp():
     return dp_attention
 
 
-def _sa():
-    from sglang.srt import server_args
-
-    return server_args
-
-
 _PARALLEL_FIELDS = frozenset(
     {
         "world_size",
@@ -223,15 +219,29 @@ class RuntimeContext:
     """Container for the structured runtime accessors; exposes ``parallel`` and
     ``server_args``."""
 
-    __slots__ = ("parallel",)
+    __slots__ = ("parallel", "_server_args")
 
     def __init__(self, parallel: ParallelContext):
         self.parallel = parallel
+        self._server_args: ServerArgs | None = None
 
     @property
     def server_args(self) -> ServerArgs:
-        """The process-wide ``ServerArgs``, read through the global getter."""
-        return _sa().get_global_server_args()
+        """The process-wide ``ServerArgs`` (context-owned slot)."""
+        server_args = self._server_args
+        if server_args is None:
+            # Verbatim legacy message: tests and user scripts may match on it.
+            raise ValueError("Global server args is not set yet!")
+        return server_args
+
+    def set_server_args(self, server_args: ServerArgs) -> None:
+        """Publish the process-wide ``ServerArgs`` into the context-owned slot.
+
+        Overwrite-allowed: a re-publish replaces the slot (test kits re-publish
+        per test; production ordering discipline lives at the call-sites, e.g.
+        the draft-worker guard in ``ModelRunner.__init__``).
+        """
+        self._server_args = server_args
 
 
 _PARALLEL = ParallelContext()
@@ -248,3 +258,11 @@ def get_parallel() -> ParallelContext:
 
 def get_server_args() -> ServerArgs:
     return _CONTEXT.server_args
+
+
+def reset_context() -> None:
+    """Clear the context-owned store (unit-test teardown).
+
+    Wrapper subsystems (``parallel``) hold no state and are unaffected.
+    """
+    _CONTEXT._server_args = None

python/sglang/srt/server_args.py

@@ -7594,23 +7594,23 @@ class ServerArgs:
         }
 
 
-# NOTE: This is a global variable to hold the server args for scheduler.
-_global_server_args: Optional[ServerArgs] = None
-
-
+# NOTE: The process-wide ServerArgs is owned by the runtime context
+# (sglang.srt.runtime_context). The two functions below are thin shims kept for
+# the existing call-sites; they publish/read the same live object by reference.
+# Imports are in-function so the two modules stay cycle-free at import time.
 def set_global_server_args_for_scheduler(server_args: ServerArgs):
-    global _global_server_args
-    _global_server_args = server_args
+    from sglang.srt.runtime_context import get_context
+
+    get_context().set_server_args(server_args)
 
 
 set_global_server_args_for_tokenizer = set_global_server_args_for_scheduler
 
 
 def get_global_server_args() -> ServerArgs:
-    if _global_server_args is None:
-        raise ValueError("Global server args is not set yet!")
+    from sglang.srt.runtime_context import get_context
 
-    return _global_server_args
+    return get_context().server_args
 
 
 def prepare_server_args(argv: List[str]) -> ServerArgs:

test/registered/unit/test_runtime_context.py

@@ -7,18 +7,19 @@ register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 import unittest
 from unittest.mock import patch
 
+import sglang.srt.server_args as server_args_module
 from sglang.srt.runtime_context import (
     ParallelContext,
     RuntimeContext,
     get_context,
     get_parallel,
     get_server_args,
+    reset_context,
 )
 from sglang.test.test_utils import CustomTestCase
 
 _PS = "sglang.srt.distributed.parallel_state"
 _DP = "sglang.srt.layers.dp_attention"
-_SA = "sglang.srt.server_args"
 
 SIZE_RANK_DELEGATIONS = [
     ("world_size", f"{_PS}.get_world_size"),
@@ -144,41 +145,67 @@ class TestParallelOverride(_IsolatedOverrides):
         self.assertEqual(p._overrides, {})
 
 
-class TestServerArgsReadThrough(CustomTestCase):
-    """``server_args`` delegates live to the global getter (read-through, V2a)."""
+class _IsolatedServerArgs(CustomTestCase):
+    """Save/restore the published ServerArgs around each test (the slot is
+    process-global; another test file sharing the process may have published)."""
 
-    def test_delegates_to_global_getter(self):
+    def setUp(self):
+        super().setUp()
+        self._saved_server_args = get_context()._server_args
+
+    def tearDown(self):
+        if self._saved_server_args is None:
+            reset_context()
+        else:
+            get_context().set_server_args(self._saved_server_args)
+        super().tearDown()
+
+
+class TestServerArgsOwnership(_IsolatedServerArgs):
+    """V2b: the context owns the slot; the legacy getters are identity shims."""
+
+    def test_legacy_setter_publishes_into_context(self):
+        # Identity (not equality) is the contract; publish accepts any object.
         sentinel = object()
-        with patch(f"{_SA}.get_global_server_args", return_value=sentinel):
+        server_args_module.set_global_server_args_for_scheduler(sentinel)
+        self.assertIs(server_args_module.get_global_server_args(), sentinel)
         self.assertIs(get_server_args(), sentinel)
         self.assertIs(get_context().server_args, sentinel)
 
-    def test_identity_with_global_getter(self):
-        import sglang.srt.server_args as server_args_module
-
-        # Identity (not equality) is the contract; publish accepts any object.
+    def test_context_publish_visible_through_legacy_getter(self):
         sentinel = object()
-        saved = server_args_module._global_server_args
-        try:
-            server_args_module.set_global_server_args_for_scheduler(sentinel)
+        get_context().set_server_args(sentinel)
+        self.assertIs(server_args_module.get_global_server_args(), sentinel)
+
+    def test_tokenizer_alias_is_same_function(self):
         self.assertIs(
-                get_server_args(), server_args_module.get_global_server_args()
+            server_args_module.set_global_server_args_for_tokenizer,
+            server_args_module.set_global_server_args_for_scheduler,
         )
-            self.assertIs(get_server_args(), sentinel)
-        finally:
-            server_args_module._global_server_args = saved
 
-    def test_pre_publish_error_passes_through(self):
-        import sglang.srt.server_args as server_args_module
-
-        saved = server_args_module._global_server_args
-        server_args_module._global_server_args = None
-        try:
+    def test_pre_publish_error_verbatim(self):
+        reset_context()
+        for accessor in (get_server_args, server_args_module.get_global_server_args):
             with self.assertRaises(ValueError) as cm:
-                get_server_args()
+                accessor()
             self.assertEqual(str(cm.exception), "Global server args is not set yet!")
-        finally:
-            server_args_module._global_server_args = saved
+
+    def test_republish_overwrite_allowed(self):
+        first, second = object(), object()
+        server_args_module.set_global_server_args_for_scheduler(first)
+        server_args_module.set_global_server_args_for_scheduler(second)
+        self.assertIs(get_server_args(), second)
+
+    def test_reset_context_clears_owned_store(self):
+        server_args_module.set_global_server_args_for_scheduler(object())
+        reset_context()
+        with self.assertRaises(ValueError):
+            get_server_args()
+
+    def test_module_global_removed(self):
+        # The legacy storage must not survive: a stale _global_server_args would
+        # silently fork the config into two objects.
+        self.assertFalse(hasattr(server_args_module, "_global_server_args"))
 
 
 if __name__ == "__main__":