From 1b15ef775ca4ab1e175d59c816250079de27c888 Mon Sep 17 00:00:00 2001 From: Hanwen Date: Mon, 24 Mar 2025 06:42:38 -0700 Subject: [PATCH] [integ-tests] Reduce timeout limit The boot speed of Rocky and RHEL were improved after https://github.com/aws/aws-parallelcluster-cookbook/pull/2920 Signed-off-by: Hanwen --- tests/integration-tests/tests/common/hit_common.py | 4 +--- tests/integration-tests/tests/schedulers/test_slurm.py | 4 +--- 2 files changed, 2 insertions(+), 6 deletions(-) diff --git a/tests/integration-tests/tests/common/hit_common.py b/tests/integration-tests/tests/common/hit_common.py index b3bc1b8b6d..dc8165c037 100644 --- a/tests/integration-tests/tests/common/hit_common.py +++ b/tests/integration-tests/tests/common/hit_common.py @@ -85,9 +85,7 @@ def assert_compute_node_reasons(scheduler_commands, compute_nodes, expected_reas assert_that(node_info).contains(f"Reason={expected_reason}") -# TOFIX We observe in 3.13.0 an increase in the bootstrap time for Rocky and RHEL. -# We must address it and restore the default wait time to 5 minutes. -@retry(wait_fixed=seconds(20), stop_max_delay=minutes(7)) +@retry(wait_fixed=seconds(20), stop_max_delay=minutes(5)) def wait_for_num_nodes_in_scheduler(scheduler_commands, desired, filter_by_partition=None): assert_num_nodes_in_scheduler(scheduler_commands, desired, filter_by_partition) diff --git a/tests/integration-tests/tests/schedulers/test_slurm.py b/tests/integration-tests/tests/schedulers/test_slurm.py index 41c2ed9ba0..8f1010164c 100644 --- a/tests/integration-tests/tests/schedulers/test_slurm.py +++ b/tests/integration-tests/tests/schedulers/test_slurm.py @@ -257,9 +257,7 @@ def test_slurm_scaling( cluster = clusters_factory(cluster_config) remote_command_executor = RemoteCommandExecutor(cluster) scheduler_commands = scheduler_commands_factory(remote_command_executor) - # TOFIX We observe in 3.13.0 an increase in the bootstrap time for Rocky and RHEL. - # We must address it and restore the default wait time to 300s. - stop_max_delay_secs = 400 if (os.startswith("rocky") or os.startswith("rhel")) else 300 + stop_max_delay_secs = 300 _assert_cluster_initial_conditions(scheduler_commands, 20, 20, 4) _test_online_node_configured_correctly(