[Perf] Enable environment cache in EngineCore to enable the feature for...

[Perf] Enable environment cache in EngineCore to enable the feature for UniProcExecutor as well (#29289)
Signed-off-by: Jialin Ouyang <Jialin.Ouyang@gmail.com>

tests/test_envs.py

@@ -8,6 +8,7 @@ import pytest
 
 import vllm.envs as envs
 from vllm.envs import (
+    disable_envs_cache,
     enable_envs_cache,
     env_list_with_choices,
     env_set_with_choices,
@@ -57,6 +58,43 @@ def test_getattr_with_cache(monkeypatch: pytest.MonkeyPatch):
     envs.__getattr__ = envs.__getattr__.__wrapped__
 
 
+def test_getattr_with_reset(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setenv("VLLM_HOST_IP", "1.1.1.1")
+    # __getattr__ is not decorated with functools.cache
+    assert not hasattr(envs.__getattr__, "cache_info")
+
+    # Enable envs cache and ignore ongoing environment changes
+    enable_envs_cache()
+    assert envs.VLLM_HOST_IP == "1.1.1.1"
+    # With cache enabled, the environment variable value is cached and unchanged
+    monkeypatch.setenv("VLLM_HOST_IP", "2.2.2.2")
+    assert envs.VLLM_HOST_IP == "1.1.1.1"
+
+    disable_envs_cache()
+    assert envs.VLLM_HOST_IP == "2.2.2.2"
+    # After cache disabled, the environment variable value would be synced
+    # with os.environ
+    monkeypatch.setenv("VLLM_HOST_IP", "3.3.3.3")
+    assert envs.VLLM_HOST_IP == "3.3.3.3"
+
+
+def test_is_envs_cache_enabled() -> None:
+    assert not envs._is_envs_cache_enabled()
+    enable_envs_cache()
+    assert envs._is_envs_cache_enabled()
+
+    # Only wrap one-layer of cache, so we only need to
+    # call disable once to reset.
+    enable_envs_cache()
+    enable_envs_cache()
+    enable_envs_cache()
+    disable_envs_cache()
+    assert not envs._is_envs_cache_enabled()
+
+    disable_envs_cache()
+    assert not envs._is_envs_cache_enabled()
+
+
 class TestEnvWithChoices:
     """Test cases for env_with_choices function."""
 

vllm/distributed/parallel_state.py

@@ -1586,6 +1586,8 @@ def destroy_distributed_environment():
 
 
 def cleanup_dist_env_and_memory(shutdown_ray: bool = False):
+    # Reset environment variable cache
+    envs.disable_envs_cache()
     # Ensure all objects are not frozen before cleanup
     gc.unfreeze()
 

vllm/envs.py

@@ -1580,6 +1580,12 @@ def __getattr__(name: str):
     raise AttributeError(f"module {__name__!r} has no attribute {name!r}")
 
 
+def _is_envs_cache_enabled() -> bool:
+    """Checked if __getattr__ is wrapped with functools.cache"""
+    global __getattr__
+    return hasattr(__getattr__, "cache_clear")
+
+
 def enable_envs_cache() -> None:
     """
     Enables caching of environment variables. This is useful for performance
@@ -1590,6 +1596,9 @@ def enable_envs_cache() -> None:
     runtime overhead. This also means that environment variables should NOT
     be updated after the service is initialized.
     """
+    if _is_envs_cache_enabled():
+        # Avoid wrapping functools.cache multiple times
+        return
     # Tag __getattr__ with functools.cache
     global __getattr__
     __getattr__ = functools.cache(__getattr__)
@@ -1599,6 +1608,17 @@ def enable_envs_cache() -> None:
         __getattr__(key)
 
 
+def disable_envs_cache() -> None:
+    """
+    Resets the environment variables cache. It could be used to isolate environments
+    between unit tests.
+    """
+    global __getattr__
+    # If __getattr__ is wrapped by functions.cache, unwrap the caching layer.
+    if _is_envs_cache_enabled():
+        __getattr__ = __getattr__.__wrapped__
+
+
 def __dir__():
     return list(environment_variables.keys())
 

vllm/v1/engine/core.py

@@ -211,6 +211,9 @@ class EngineCore:
         freeze_gc_heap()
         # If enable, attach GC debugger after static variable freeze.
         maybe_attach_gc_debug_callback()
+        # Enable environment variable cache (e.g. assume no more
+        # environment variable overrides after this point)
+        enable_envs_cache()
 
     def _initialize_kv_caches(
         self, vllm_config: VllmConfig
@@ -672,10 +675,6 @@ class EngineCoreProc(EngineCore):
                 assert addresses.coordinator_input is not None
                 logger.info("Waiting for READY message from DP Coordinator...")
 
-        # Enable environment variable cache (e.g. assume no more
-        # environment variable overrides after this point)
-        enable_envs_cache()
-
     @contextmanager
     def _perform_handshakes(
         self,