From 20af9cef17374a287850a866983943bcad579fa2 Mon Sep 17 00:00:00 2001 From: "Alex Chi Z." <4198311+skyzh@users.noreply.github.com> Date: Tue, 4 Mar 2025 09:55:50 -0500 Subject: [PATCH] fix(test): use the same value for reldir v1+v2 (#11070) ## Problem part of https://github.com/neondatabase/neon/issues/11067 My observation is that with the current value of settings, x86-v1 usually takes 30s, arm-v1 1m30s, x86-v2 1m, arm-v2 3m. But sometimes the system could run too slow and cause test to timeout on arm with reldir v2. While I investigate what's going on and further improve the performance, I'd like to set both of them to use the same test input, so that it doesn't timeout and we don't abuse this test case as a performance test. ## Summary of changes Use the same settings for both test cases. Signed-off-by: Alex Chi Z --- test_runner/regress/test_pg_regress.py | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-) diff --git a/test_runner/regress/test_pg_regress.py b/test_runner/regress/test_pg_regress.py index df243c13f1..d2a78b16e4 100644 --- a/test_runner/regress/test_pg_regress.py +++ b/test_runner/regress/test_pg_regress.py @@ -375,12 +375,8 @@ def test_tx_abort_with_many_relations( # How many relations: this number is tuned to be long enough to take tens of seconds # if the rollback code path is buggy, tripping the test's timeout. - if reldir_type == "v1": - n = 4000 - step = 4000 - else: - n = 20000 - step = 5000 + n = 5000 + step = 2500 def create(): # Create many relations