fix(PD): respect pause_generation in disagg event loops (#20908)

python/sglang/srt/disaggregation/decode.py

@@ -1178,8 +1178,9 @@ class SchedulerDisaggregationDecodeMixin:
             # Receive requests
             recv_reqs = self.recv_requests()
             self.process_input_requests(recv_reqs)
-            # polling and allocating kv cache
             self.process_decode_queue()
+            if self._engine_paused:
+                continue
 
             # Get the next batch to run
             batch = self.get_next_disagg_decode_batch_to_run()
@@ -1205,8 +1206,9 @@ class SchedulerDisaggregationDecodeMixin:
             # Receive requests
             recv_reqs = self.recv_requests()
             self.process_input_requests(recv_reqs)
-            # polling and allocating kv cache
             self.process_decode_queue()
+            if self._engine_paused:
+                continue
 
             # Get the next batch to run
             batch = self.get_next_disagg_decode_batch_to_run()

python/sglang/srt/disaggregation/prefill.py

@@ -397,6 +397,9 @@ class SchedulerDisaggregationPrefillMixin:
             self.waiting_queue.extend(
                 self.disagg_prefill_bootstrap_queue.pop_bootstrapped()
             )
+            if self._engine_paused:
+                self.process_disagg_prefill_inflight_queue()
+                continue
 
             # Get the next batch to run
             batch = self.get_next_disagg_prefill_batch_to_run()
@@ -428,6 +431,9 @@ class SchedulerDisaggregationPrefillMixin:
             self.waiting_queue.extend(
                 self.disagg_prefill_bootstrap_queue.pop_bootstrapped()
             )
+            if self._engine_paused:
+                self.process_disagg_prefill_inflight_queue()
+                continue
 
             # Get the next batch to run
             batch = self.get_next_disagg_prefill_batch_to_run()