xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 04:54:56 +08:00
Code Issues Packages Projects Releases Wiki Activity

[P/D] Log warnings related to prefill KV expiry (#21753)

Browse Source 
Signed-off-by: Nick Hill <nhill@redhat.com>
This commit is contained in:
Nick Hill 2025-07-28 19:40:53 +01:00 committed by GitHub
parent e17a4d3bf9
commit 7d44c691b0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 12 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
View File

@ -1025,6 +1025,11 @@ class NixlConnectorWorker:
# Sorted dict, oldest requests are put first so we can exit early.
if now < expires:
break
count = self.consumer_notification_counts_by_req.pop(req_id, 0)
logger.warning(
"Releasing expired KV blocks for request %s which were "
"retrieved by %d decode worker(s) within %d seconds.", req_id,
count, envs.VLLM_NIXL_ABORT_REQUEST_TIMEOUT)
del self._reqs_to_send[req_id]
done_sending.add(req_id)
@ -1040,6 +1045,13 @@ class NixlConnectorWorker:
for notifs in self.nixl_wrapper.get_new_notifs().values():
for notif in notifs:
req_id, tp_ratio = notif.decode("utf-8").rsplit(":", 1)
if req_id not in self._reqs_to_send:
logger.error(
"Potentially invalid KV blocks for "
"unrecognized request %s were retrieved by "
"a decode worker. They may have expired.", req_id)
continue
self.consumer_notification_counts_by_req[req_id] += 1
# Wait all consumers (D) to be done reading before freeing.
if self.consumer_notification_counts_by_req[req_id] == int(