[BugFix] [KVConnector] Fix KV events for LMCache connector

[hickeyma]

Purpose

A small fix for the LMCache connector as LMCache does not yet support lora_name as an events property. Bug introduced in #27577.

The error occurs when try to prompt a model with LMCache connector and KV events enabled:

[...]
(EngineCore_DP0 pid=223048) [2026-01-12 11:02:52,233] LMCache INFO: Reqid: cmpl-a310c95f98e4f404-0-8bf3d3d5, Total tokens 32, LMCache hit tokens: 0, need to load: 0 (vllm_v1_adapter.py:1602:lmcache.integration.vllm.vllm_v1_adapter)
(Worker_TP1 pid=223151) [2026-01-12 11:02:52,271] LMCache INFO: Storing KV cache for 32 out of 32 tokens (skip_leading_tokens=0) for request cmpl-a310c95f98e4f404-0-8bf3d3d5 (vllm_v1_adapter.py:1374:lmcache.integration.vllm.vllm_v1_adapter)
(Worker_TP0 pid=223150) [2026-01-12 11:02:52,271] LMCache INFO: Storing KV cache for 32 out of 32 tokens (skip_leading_tokens=0) for request cmpl-a310c95f98e4f404-0-8bf3d3d5 (vllm_v1_adapter.py:1374:lmcache.integration.vllm.vllm_v1_adapter)
(Worker_TP1 pid=223151) [2026-01-12 11:02:52,273] LMCache INFO: Stored 32 out of total 32 tokens. size: 0.0024 gb, cost 1.8921 ms, throughput: 1.2903 GB/s; offload_time: 1.8005 ms, put_time: 0.0916 ms (cache_engine.py:446:lmcache.v1.cache_engine)
(Worker_TP0 pid=223150) [2026-01-12 11:02:52,274] LMCache INFO: Stored 32 out of total 32 tokens. size: 0.0024 gb, cost 2.0154 ms, throughput: 1.2113 GB/s; offload_time: 1.9142 ms, put_time: 0.1013 ms (cache_engine.py:446:lmcache.v1.cache_engine)
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822] WorkerProc hit an exception.
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822] Traceback (most recent call last):
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]   File "/data/dev/github.com/vllm-project/vllm/vllm/v1/executor/multiproc_executor.py", line 817, in worker_busy_loop
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]     output = func(*args, **kwargs)
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]              ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]   File "/data/dev/github.com/vllm-project/vllm/vllm/v1/worker/worker_base.py", line 365, in execute_model
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]     return self.worker.execute_model(scheduler_output)
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]   File "/data/conda-envs/vllm-test/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]     return func(*args, **kwargs)
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]            ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]   File "/data/dev/github.com/vllm-project/vllm/vllm/v1/worker/gpu_worker.py", line 629, in execute_model
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]     output = self.model_runner.execute_model(
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]   File "/data/conda-envs/vllm-test/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]     return func(*args, **kwargs)
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]            ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]   File "/data/dev/github.com/vllm-project/vllm/vllm/v1/worker/gpu_model_runner.py", line 3277, in execute_model
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]     self.maybe_get_kv_connector_output(scheduler_output) as kv_connector_output,
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]   File "/data/conda-envs/vllm-test/lib/python3.12/contextlib.py", line 144, in __exit__
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]     next(self.gen)
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]   File "/data/dev/github.com/vllm-project/vllm/vllm/v1/worker/kv_connector_model_runner_mixin.py", line 139, in _get_kv_connector_output
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]     output.kv_cache_events = kv_connector.get_kv_connector_kv_cache_events()
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]   File "/data/dev/github.com/vllm-project/vllm/vllm/distributed/kv_transfer/kv_connector/v1/lmcache_connector.py", line 237, in get_kv_connector_kv_cache_events
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]     lora_name=e.lora_name,
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822]               ^^^^^^^^^^^
(Worker_TP1 pid=223151) ERROR 01-12 11:02:52 [multiproc_executor.py:822] AttributeError: 'CacheStoreEvent' object has no attribute 'lora_name'

Test Plan

Test Result

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

---

Note

Prevents crashes when using LMCache KV events by aligning with current LMCache event schema.

• In lmcache_connector.py, set lora_name=None when creating BlockStored from LMCache events (LMCache does not provide lora_name yet).

^{Written by Cursor Bugbot for commit 609dec1. This will update automatically on new commits. Configure here.}

---

Note

Prevents crashes when LMCache KV events omit lora_name.

• In lmcache_connector.py, construct BlockStored with lora_name=getattr(e, "lora_name", None) to align with current LMCache event schema and maintain compatibility

^{Written by Cursor Bugbot for commit ecaab07. This will update automatically on new commits. Configure here.}

---

Note

^{Cursor Bugbot is generating a summary for commit 0233a46. Configure here.}

---

Note

Fixes LMCache KV events compatibility by making lora_name optional when constructing BlockStored.

• In lmcache_connector.py, set lora_name=getattr(e, "lora_name", None) to align with LMCache events that omit lora_name, preventing AttributeError crashes during KV event processing

^{Written by Cursor Bugbot for commit 0233a46. This will update automatically on new commits. Configure here.}

[gemini-code-assist]

Code Review

This pull request addresses a critical AttributeError that occurs when using the LMCache connector with KV events enabled, due to the lora_name property not being available in the current version of LMCache. The fix correctly resolves the crash by setting lora_name to None. I've provided one suggestion to make the fix more robust and future-proof by using getattr, which will automatically adapt when LMCache adds support for lora_name in a future release.

[NickLucche]

Thanks @hickeyma !