[PD] Add the missing Prefill bootstrap timeout for NIXL (#34692)
This commit is contained in:
@@ -2745,6 +2745,7 @@ class NixlKVSender(CommonKVSender):
|
|||||||
pp_rank,
|
pp_rank,
|
||||||
req_has_disagg_prefill_dp_rank,
|
req_has_disagg_prefill_dp_rank,
|
||||||
)
|
)
|
||||||
|
self.init_time = time.time()
|
||||||
self.has_sent = False
|
self.has_sent = False
|
||||||
self.chunk_id = 0
|
self.chunk_id = 0
|
||||||
self._send_failed = False
|
self._send_failed = False
|
||||||
@@ -2790,6 +2791,10 @@ class NixlKVSender(CommonKVSender):
|
|||||||
if self._send_failed:
|
if self._send_failed:
|
||||||
return KVPoll.Failed # type: ignore
|
return KVPoll.Failed # type: ignore
|
||||||
status = self.kv_mgr.check_status(self.bootstrap_room)
|
status = self.kv_mgr.check_status(self.bootstrap_room)
|
||||||
|
if status == KVPoll.Bootstrapping:
|
||||||
|
timeout_result = self._check_bootstrap_timeout()
|
||||||
|
if timeout_result is not None:
|
||||||
|
return timeout_result
|
||||||
# Hold Success until all staging chunks transferred: a deferred chunk
|
# Hold Success until all staging chunks transferred: a deferred chunk
|
||||||
# can still be pending, and concluding now would drop it.
|
# can still be pending, and concluding now would drop it.
|
||||||
if (
|
if (
|
||||||
|
|||||||
Reference in New Issue
Block a user