[misc] Remove break-graph debug log; reclaim pid-less /dev/shm leaks in CI (#33929)

This commit is contained in:
Liangsheng Yin
2026-08-06 21:18:01 -07:00
committed by GitHub
parent 0c3a76fa0a
commit afa79330b8
3 changed files with 51 additions and 22 deletions
@@ -44,13 +44,13 @@ class TestCleanupStaleShm(unittest.TestCase):
def _make_segment(self, name: str) -> str:
shm = shared_memory.SharedMemory(create=True, size=4096, name=name)
shm.close()
self.addCleanup(self._unlink_quiet, name)
self.addCleanup(self._unlink_quiet, f"/dev/shm/{name}")
return name
@staticmethod
def _unlink_quiet(name: str):
def _unlink_quiet(path: str):
try:
shared_memory.SharedMemory(name=name).unlink()
os.unlink(path)
except FileNotFoundError:
pass
@@ -121,6 +121,33 @@ class TestCleanupStaleShm(unittest.TestCase):
self.assertFalse(os.path.exists(f"/dev/shm/{stale}"))
def _make_raw_file(self, name: str) -> str:
"""Orphan families are plain files, not shared_memory segments."""
path = f"/dev/shm/{name}"
with open(path, "wb") as f:
f.write(b"\0" * 4096)
self.addCleanup(self._unlink_quiet, path)
return path
@unittest.skipUnless(
os.environ.get("SGLANG_IS_IN_CI", "").lower() in ("true", "1"),
"sweeps orphan families unconditionally; only safe on a CI runner",
)
def test_orphan_family_sweep(self):
stale = [
self._make_raw_file("sglang_loads_test_deadbeef.shm"),
self._make_raw_file("cuda.shm.0.deadbeef.1"),
self._make_raw_file("nccl-testonly"),
self._make_raw_file("sem.loky-0-testonly"),
]
unknown = self._make_raw_file("unknown_family_file")
cleanup_stale_shm()
for path in stale:
self.assertFalse(os.path.exists(path), path)
self.assertTrue(os.path.exists(unknown))
if __name__ == "__main__":
unittest.main()