slurm - 通过 gres 对 SLURM 报告进行排序不起作用

Question

我想查看我正在管理的 SLURM 集群上用户的 gres 使用情况。问题是我无法根据这种用法对其进行排序，它按 cpu 使用情况（或计费？）根据https://slurm.schedmd.com/sreport.html TopUsage 进行排序“......用户按 CPUTime 排序, 但 -T, --tres 选项将按指定的第一个 TRES 对用户进行排序。” 但我得到的输出是

sreport  user  -t hourper TopUsage TopCount=5 --tres=gres/gpu:tesla start=2021-05-01T00:00:00 
--------------------------------------------------------------------------------
Top 5 Users 2021-05-01T00:00:00 - 2021-09-28T23:59:59 (13046400 secs)
Usage reported in TRES Hours/Percentage of Total
--------------------------------------------------------------------------------
  Cluster     Login     Proper Name         Account      TRES Name              Used 
--------- --------- --------------- --------------- -------------- ----------------- 
    bayes   a               a               student gres/gpu:tesla          0(0.00%) 
    bayes   b               b               student gres/gpu:tesla          0(0.00%) 
    bayes   c               c               student gres/gpu:tesla          0(0.00%) 
    bayes   d               d               student gres/gpu:tesla          2(0.05%) 
    bayes   e               e               student gres/gpu:tesla          0(0.00%) 

如果我还将 cpu 添加到我得到的 tres

sreport  user  -t hourper TopUsage TopCount=5 --tres=gres/gpu:tesla,cpu start=2021-05-01T00:00:00 
--------------------------------------------------------------------------------
Top 5 Users 2021-05-01T00:00:00 - 2021-09-28T23:59:59 (13046400 secs)
Usage reported in TRES Hours/Percentage of Total
--------------------------------------------------------------------------------
  Cluster     Login     Proper Name         Account      TRES Name              Used 
--------- --------- --------------- --------------- -------------- ----------------- 
    bayes         a               a         student            cpu      13786(1.59%) 
    bayes         a               a         student gres/gpu:tesla          0(0.00%) 
    bayes         b               b         student            cpu      13509(1.55%) 
    bayes         b               b         student gres/gpu:tesla          0(0.00%) 
    bayes         c               c         student            cpu       6134(0.71%) 
    bayes         c               c         student gres/gpu:tesla          0(0.00%) 
    bayes         d               d         student            cpu       2792(0.32%) 
    bayes         d               d         student gres/gpu:tesla          2(0.05%) 
    bayes         e               e         student            cpu       1585(0.18%) 
    bayes         e               e         student gres/gpu:tesla          0(0.00%)

关于如何正确排序的任何想法？

标签： slurm