[slurm-dev] Re: cpu identifier

Squee.(*) Just a note - for Mellanox IB users, there is the tuning guide which advises using interrupts on the CPU nearest the HBA. I guess it makes sense to eke out that last fraction of performance to make the reserved cores be local to the HBA. hwloc is you friend here. (*) and see my

[slurm-dev] Re: cpu identifier

Andrealphus,# You should be using cpusets You allocate cores 1 and 2 (actually I think they count from 0) as the 'boot cpuset' and run the operating system processes in that. You then create a cpuset for each job. I have done this with PBSPro and it works very well. http://slurm.schedmd.com/cgro

[slurm-dev] Re: cpu identifier

Thanks Christopher! On Wed, Sep 14, 2016 at 4:30 PM, Christopher Samuel wrote: > > On 15/09/16 05:20, andrealphus wrote: > >> On a side note, any idea if there is a parameter to not >> have it use a particular cpu? This is a single node workstation, with >> 18 cores. The end goal is to have a de

[slurm-dev] Re: cpu identifier

On 15/09/16 05:20, andrealphus wrote: > On a side note, any idea if there is a parameter to not > have it use a particular cpu? This is a single node workstation, with > 18 cores. The end goal is to have a default set up where it can say > run 16 jobs, bound to 16 unique cores, excluding core 1 a

[slurm-dev] Task Plugin Configuration

What is the best way to make a configuration file for a task plugin? I just want the user to be able to define some directories in a configuration file, then in the plugin I want to loop through those directory variables to run by plugin code on them. Does anyone know of an example plugin I can loo

[slurm-dev] multiple login nodes

Hi all We have a cluster with two login nodes. login1 installed with rhel6 and login2 with rhel7. The slurmctl run in master node with rhel6. All very good, but in the nodes with rhel6 exists modules-file, this modules-file no exist in nodes with rhel7. I can send jobs to partitions whith rhel7 f

[slurm-dev] Re: cpu identifier

Thanks Dani! On a side note, any idea if there is a parameter to not have it use a particular cpu? This is a single node workstation, with 18 cores. The end goal is to have a default set up where it can say run 16 jobs, bound to 16 unique cores, excluding core 1 and 2, which are primarily used fo

[slurm-dev] Prolog script (maybe) question?

Hi there, I think it ought to be possible to use a Prolog script to look over a user's submission and to check, for example, if the user has chosen to work/write output into their home directory as opposed to a faster scratch area. Does anyone have an example of such a script, or a script that

[slurm-dev] Job accounting

We've been using the multifactor plugin to enforce accounting with grpcpumin limits for accounts. That's worked well with a homogeneous node configuration. We've historically allocated time by node-hours, but with all nodes being identical, it's pretty easy to do a uniform conversion from cpu

[slurm-dev] differentiate between interactive and batch jobs in job_submit.lua?

Hello everyone, is it possible to check in the submit lua script whether a job has been submitted with sbatch or salloc? The batch_flag is apparently not present in the job_descriptor struct. Kind Regards, Lech

[slurm-dev] Re: Want to contribute, where should I start?

Want to contribute, where should I start? Hi Felipe, I point to these comments whenever someone asks that question :-) From 1: o�� tinkerghost ��� I recommend starting by

[slurm-dev] Problem with power saving (slurm-16.05.4)

Hi, The patch introduced in commit b4904661 (not powering down dead node) has a bad side effect. Adding the "(node_ptr->last_idle != 0)" condition prevents from powering down nodes with the following command: scontrol update nodename=nX state=power_down because the state update function rel

[slurm-dev] Re: cpu identifier

Is there binding/affinity involved? If not, the process might execute each instruction on a different cpu, so slurm couldn't really provide something useful. If there is binding, take a look at SBATCH_CPU_BIND Set to value of the --cpu_bind option. SBATCH_CPU_BIND_VERBOSE Set to "ver