英文:
Detremine total core hours to be billed for a slurm job?
问题
以下是您要翻译的内容：
我有一个关于HPC计费的基本问题。我浏览了网站，但找不到明确的答案。
对于给定的作业ID，如附图所示，我似乎有三个正在运行的作业。我使用一个sbatch命令运行一个bash脚本，其中包括一个srun命令。我认为第2行是bash脚本，第3行是srun命令。我想知道我将为这个作业被计费的总CPU小时数是多少？它是否为
- 3591397+3591397.bat+3591397.0 ～ 5小时
- 3591397.bat+3591397.0 ～ 3小时13分钟
特定作业的sacct结果
我知道还有其他因素，如内存、使用的GPU，可能会导致总计费发生变化，因为它们具有不同于1的计费权重。但是，我现在不想考虑它们。
英文:
I have a basic question about HPC billing. I’ve surfed the site but could not find a clear answer.
For a given job ID, I have seemingly three jobs running as shown in the attached photo. I run a bash script with one srun command using sbatch command. I believe the 2nd line is for the bash script and the 3rd is for the srun command. I want to know the total CPU hours I'll be billed for this job? Would it be the time for
- 3591397+3591397.bat+3591397.0 ~ 5 hrs
- 3591397.bat+3591397.0 ~ 3:13 hrs
results for the sacct for a particular job
I know there are other considerations like memory, GPUs used that can change the total billing by having a billing weight different than 1. However, I don’t want to consider them for now.
Thanks.
答案1
得分: 0
只有一个与作业ID 3591397相关联的作业，但该作业有两个步骤，一个是3591397.batch，对应于提交脚本，另一个是3591397.0，对应于提交脚本中的第一个（似乎是唯一的）srun 实例。
第一行是作业的摘要，因此计费将为13分钟49秒乘以8。
在您的输出中，.batch 步骤的“已用时间（Elapsed）”值之间存在一秒的差异，但这可能是由于四舍五入误差引起的。
如果您不对各个步骤的信息感兴趣，可以使用sacct的-X参数，只获取每个作业的一行（在您的示例中是第一行）。
英文:
You only have one job associated to that job ID 3591397, but that job has two steps, one 3591397.batch that corresponds to the submission script, and one 3591397.0 that corresponds to the first (and seemingly only) instance of srun in the submission script.
The first line is the summary for the job so the billing would be 13 minutes and 49 seconds multiplied by 8.
In your output, there is a one-second discrepancy between the Elapsed values for the .batch step, but that could be due to rounding errors.
If you are not interested in the information for individual steps, you can use the -X parameter of sacct to only get one line per job (the first line in your example).
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论