xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 08:04:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

[P/D] Avoid stranding blocks in P when aborted in D's waiting queue (#19223)

Browse Source 
Signed-off-by: Nick Hill <nhill@redhat.com>
This commit is contained in:
Nick Hill 2025-06-25 15:19:44 -07:00 committed by GitHub
parent 2cc2069970
commit 55c65ab495
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 14 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
View File

@ -298,8 +298,21 @@ class NixlConnectorScheduler:
logger.debug(
"NIXLConnector request_finished, request_status=%s, "
"kv_transfer_params=%s", request.status, params)
if not params:
return False, None
if (params is None or not params.get("do_remote_decode")
if params.get("do_remote_prefill"):
# If do_remote_prefill is still True when the request is finished,
# update_state_after_alloc must not have been called (the request
# must have been aborted before it was scheduled).
# To avoid stranding the prefill blocks in the prefill instance,
# we must add empty block_ids to _reqs_need_recv so that our
# worker side will notify and free blocks in the prefill instance.
self._reqs_need_recv[request.request_id] = (request, [])
params["do_remote_prefill"] = False
return False, None
if (not params.get("do_remote_decode")
or request.status != RequestStatus.FINISHED_LENGTH_CAPPED):
return False, None