about summary refs log tree commit diff
path: root/test/python/server_test_case.py
diff options
context:
space:
mode:
authorPjotr <wrk@bosa.lan>2026-11-05 11:51:13 +0000
committerPjotr <wrk@bosa.lan>2026-11-05 11:51:13 +0000
commit8939c00462dd96ee80052a633d6d189f74bedecf (patch)
tree34148b6a60f5db3674987d20c94df678faf116b1 /test/python/server_test_case.py
parentff67b720f3d0c74739d8b8b657ac2e96bf171a1b (diff)
downloadgenenetwork4-8939c00462dd96ee80052a633d6d189f74bedecf.tar.gz
docs: what the slurm backend costs, and why job records must stay
Measured the same request (BXD fixtures, three chromosomes) through
/compute/gemma, ravanan's single-machine backend and ravanan on Slurm,
cold and warm. The Slurm backend is dispatch-bound at this size -- one
scheduler tick per step, with each job itself running in under a second
-- so cold and warm are identical there, and the README's 16.7s figure
belongs to the single-machine backend.

Also document the Slurm-side requirement behind that measurement: a
completed job has to stay queryable for as long as the workflow runs.
Once slurmctld purges the record ravanan falls back to slurmdbd, and on
a cluster without accounting the workflow aborts on a 404 long after the
job succeeded.
Diffstat (limited to 'test/python/server_test_case.py')
0 files changed, 0 insertions, 0 deletions