Skip to content

Commit

Permalink
chore: Increase the timeout for e2e_slurm_restart tests (FOUNDENG-470) (
Browse files Browse the repository at this point in the history
#714)

The _test_master_restart_cmd tests are failing when run as part of the e2e_slurm_restart job inside the nightly workflow. The reason they are failing is that these tests pull an image as part of the test which takes more than a couple of minutes. This delay makes the experiment stay in PULLING state instead of RUNNING state and makes the check for RUNNING status inside _test_master_restart_cmd fail. I increased the timeout from 30 seconds to 300 seconds and tested the e2e_slurm_restart job manually to ensure all the tests run successfully.
  • Loading branch information
jagadeesh545 authored and hanyucui committed May 31, 2023
1 parent bd01579 commit a414e09
Showing 1 changed file with 3 additions and 1 deletion.
4 changes: 3 additions & 1 deletion e2e_tests/tests/cluster/test_master_restart.py
Original file line number Diff line number Diff line change
Expand Up @@ -225,7 +225,9 @@ def test_master_restart_cmd_k8s(

def _test_master_restart_cmd(managed_cluster: Cluster, slots: int, downtime: int) -> None:
command_id = run_command(30, slots=slots)
wait_for_command_state(command_id, "RUNNING", 30)
# The timeout below is set to 300 seconds because sometimes the experiment
# remains in the PULLING state for long duration when pulling a new image.
wait_for_command_state(command_id, "RUNNING", 300)

if downtime >= 0:
managed_cluster.kill_master()
Expand Down

0 comments on commit a414e09

Please sign in to comment.