Retire the per-runner parallel record (#40343)

test/registered/unit/disaggregation/test_decode_queue_cleanup.py

@@ -11,7 +11,6 @@ from sglang.srt.disaggregation.decode import (
 )
 from sglang.srt.disaggregation.fake.conn import FakeKVManager, FakeKVReceiver
 from sglang.srt.disaggregation.utils import DisaggregationMode
-from sglang.srt.distributed.parallel_state_wrapper import ParallelState
 from sglang.srt.managers.schedule_batch import FINISH_ABORT
 from sglang.srt.managers.scheduler import Scheduler
 from sglang.srt.runtime_context import get_context, publish, reset_context
@@ -440,7 +439,6 @@ class TestDecodeQueueCleanup(CustomTestCase):
         scheduler.last_batch = None
         scheduler.cur_batch_for_debug = None
         scheduler.enable_overlap = False
-        scheduler.ps = ParallelState.trivial()
         scheduler.running_mbs = []
         scheduler.waiting_queue = []
         scheduler.grammar_manager = SimpleNamespace(grammar_queue=[])

test/registered/unit/disaggregation/test_pd_role_switch.py

@@ -131,7 +131,6 @@ class TestHandlePdRoleSwitch(unittest.TestCase):
     def test_rejected_when_decode_graph_headroom_is_insufficient(self):
         s = self._scheduler(DisaggregationMode.PREFILL)
         s.device = "cuda"
-        s.ps = SimpleNamespace(gpu_id=0)
         s.tp_worker.get_decode_cuda_graph_bs.return_value = []
         with patch.object(role_switch, "get_available_gpu_memory", return_value=0.5):
             out = Scheduler.handle_pd_role_switch(
@@ -151,7 +150,6 @@ class TestHandlePdRoleSwitch(unittest.TestCase):
     def test_decode_graph_headroom_allows_flip(self):
         s = self._scheduler(DisaggregationMode.PREFILL)
         s.device = "cuda"
-        s.ps = SimpleNamespace(gpu_id=0)
         s.tp_worker.get_decode_cuda_graph_bs.return_value = []
         with patch.object(role_switch, "get_available_gpu_memory", return_value=1.0):
             out = Scheduler.handle_pd_role_switch(