Re: [slurm-users] ignore gpu resources to scheduled the cpu based jobs

2020-06-15 Thread Diego Zuccato
Il 13/06/20 17:47, navin srivastava ha scritto: > Yes we have separate partitions. Some are specific to gpu having 2 nodes > with 8 gpu and another partitions are mix of both,nodes with 2 gpu and > very few nodes are without any gpu.  Maybe it's already known and obvious, but... Remember that a no

Re: [slurm-users] ignore gpu resources to scheduled the cpu based jobs

2020-06-15 Thread navin srivastava
Thanks Renfro. I will perform similar setting and let us see how it goes. Regards On Mon, Jun 15, 2020, 23:02 Renfro, Michael wrote: > So if a GPU job is submitted to a partition containing only GPU nodes, and > a non-GPU job is submitted to a partition containing at least some nodes > without

[slurm-users] The Slurm User Group Meeting (SLUG'20) goes virtual this fall

2020-06-15 Thread Tim Wickberg
The Slurm User Group Meeting (SLUG'20) this fall will be moving online, and will not be hosted in-person at Harvard as scheduled, due to uncertainty around scheduling events caused by COVID-19.[1] In lieu of an in-person meeting, SchedMD will broadcast a select set of presentations on Tuesday,

Re: [slurm-users] ignore gpu resources to scheduled the cpu based jobs

2020-06-15 Thread Renfro, Michael
So if a GPU job is submitted to a partition containing only GPU nodes, and a non-GPU job is submitted to a partition containing at least some nodes without GPUs, both jobs should be able to run. Priorities should be evaluated on a per-partition basis. I can 100% guarantee that in our HPC, pendin

Re: [slurm-users] How Nice value is decided by slurm.

2020-06-15 Thread Rodrigo Santibáñez
Hi, As far I understand, the nice value is set by the user. The scheduler modifies directly the priority value. You could use the command 'sprio' to find out the contribution of the different parameters to the final priority. Best regards El lun., 15 de junio de 2020 12:50, navin srivastava esc

[slurm-users] How Nice value is decided by slurm.

2020-06-15 Thread navin srivastava
Hi, One query about how nice value will be decided by the scheduler. our scheduling policy id FIFO + Fair tree. one user submitted 100 of jobs in different dates. what i see is the old jobs are in queue but few latest job went for the execution. when i see the nice value of the latest running jo

Re: [slurm-users] configless DNS entries

2020-06-15 Thread Steven Dick
I got configless DNS SRV records working by putting the record in the default domain in the search path for the cluster. In other words, the cluster has its own domain, and all the nodes are in it along with the SRV records.. This is the first domain in the DNS search path. On Tue, Jun 9, 2020 a