[slurm-users] Getting task distribution from environment

[Alexander Grund]

Hello,





"[n1: 2, n2: 2, n4: 1]" out of "n[1-2],n4" and "2(x2),1"

Questions:

- Is there an easier way?


Thanks, Alex

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Alexander Grund
Interdisziplinäre Anwendungsunterstützung und Koordination (IAK)

Technische Universität Dresden
Zentrum für Informationsdienste und Hochleistungsrechnen (ZIH)
Würzburger Str.35/Chemnitzer Str.50, Raum 010 01062 Dresden
Tel.: +49 (351) 463-35982
E-Mail: alexander.gr...@tu-dresden.de
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

I need to get (at least) a list with the number of tasks on each hostname for the current job step from withing each task. My current approach is to use SLURM_STEP_NODELIST  and SLURM_STEP_NUM_TASKS and expand them without changing the order. Then I match them 1:1 to get something like: - Is this valid aka does this always produce the correct result? Or can the order in those 2 variables not match? - Is it possible to get some similar listing for CPUs allocated to each task? I know of SLURM_CPUS_PER_TASK but can the number of CPUs be different per task?

smime.p7s
Description: S/MIME Cryptographic Signature