From cfd302db9b8341e78c0b460d53c9c82f7df69d7d Mon Sep 17 00:00:00 2001 From: Or Ozeri Date: Tue, 30 Sep 2025 22:53:04 +0300 Subject: [PATCH] OffloadingConnector: Fix GPU block tracking bug (#25856) Signed-off-by: Or Ozeri --- .../kv_transfer/kv_connector/v1/offloading_connector.py | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py b/vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py index c23efa604544d..6936638c7f4e0 100644 --- a/vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py +++ b/vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py @@ -278,8 +278,9 @@ class OffloadingConnectorScheduler: req, start_idx=start_block_idx, end_idx=num_blocks) store_output = self.manager.prepare_store(new_block_hashes) if store_output is None: - logger.warning("Cannot store %s blocks", num_new_blocks) - break + logger.warning("Request %s: cannot store %s blocks", req_id, + num_new_blocks) + continue self._next_stored_block_idx[req_id] = num_blocks