From 528b3076afaac5ed47da85ee28966f9fd845dd78 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Nicol=C3=B2=20Lucchesi?= Date: Mon, 2 Feb 2026 12:01:29 +0100 Subject: [PATCH] [CI][Bugfix] Fix flaky `tests/v1/kv_connector/unit/test_multi_connector.py::test_multi_example_connector_consistency` (#33555) Signed-off-by: NickLucche --- tests/v1/kv_connector/unit/test_multi_connector.py | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/tests/v1/kv_connector/unit/test_multi_connector.py b/tests/v1/kv_connector/unit/test_multi_connector.py index 783678e9c..60f4002e0 100644 --- a/tests/v1/kv_connector/unit/test_multi_connector.py +++ b/tests/v1/kv_connector/unit/test_multi_connector.py @@ -151,7 +151,8 @@ def test_multi_example_connector_consistency(): kv_transfer_config=kv_transfer_config, ) # Run generation - this should trigger saving KV cache - _ = llm.generate(PROMPTS, SAMPLING_PARAMS) + # Use a single prompt to avoid race conditions depending on the order of scheduling + _ = llm.generate(PROMPTS[0], SAMPLING_PARAMS) # --- Verification --- @@ -221,7 +222,7 @@ def test_multi_example_connector_consistency(): # Run generation again - this should trigger loading from the first # connector. - _ = llm.generate(PROMPTS, SAMPLING_PARAMS) + _ = llm.generate(PROMPTS[1], SAMPLING_PARAMS) events = get_connector_events() # get_num_new_matched_tokens will return new tokens from the first @@ -247,7 +248,7 @@ def test_multi_example_connector_consistency(): # Run generation again - this should trigger loading from the first # connector. - _ = llm.generate(PROMPTS, SAMPLING_PARAMS) + _ = llm.generate(PROMPTS[0], SAMPLING_PARAMS) events = get_connector_events() # get_num_new_matched_tokens will be called for both connectors but will