slurm 不支持--share 选项-解决方法

提交脚本含有--share 的会报错,反复安装slurm尝试都失败

sbatch slurm_script sbatch: unrecognized option '--share' Try "sbatch --help" for more information

 

结果是:slurm 自从14.0后就不支持share了,改用--oversubscribe,这样就允许一个节点同时运行多个作业。

which version of slurm support ‘’--share‘’ option

 

The --share option was available in Slurm version up to 14.03. In Slurm 14.11 and later, it has been replaced by --oversubscribe. The --share option was used to allow overcommitment of node resources by multiple jobs.

In the recent versions, you can use --oversubscribe to enable the overcommitment instead. If you're running a version older than 14.11 and --share isn't working, it's possible that the system administrator has disabled it.

Please, check your Slurm version and consult the documentation or the system administrator if you encounter any issues.

 

另外在slurm.conf 中,这个参数也要开启,要不然也不可以单节点多任务跑

SelectTypeParameters=CR_Core
# SelectTypeParameters=CR_Core and . It allows multiple jobs to share the same node as long as there are enough cores available.

 
 
posted @ 2023-08-03 15:48  千家诗  阅读(60)  评论(0编辑  收藏  举报