Hi Slurm users,
We have found the need to execute a parallel command on all nodes
running jobs belonging to a particular user.
I have made a configuration to the excellent ClusterShell tool as
documented in https://wiki.fysik.dtu.dk/niflheim/SLURM#clustershell
If you add a "slurmuser" secti
Hi ,
How to set a default memory request when use srun without --mem?
for example,
$srun --mem=40G sleep 120 , it will require 40G memory, but if I run $srun
sleep 120, it may use all node's memory.
How to setup the default srun request memory size without --mem arguments?
Many thanks,
Bill
Hi ,
I just found the way , set "DefMemPerCPU=4096" for partition in slurm.conf
It will use 4G memory request.
Regards,
Bill
-- Original --
From: "Bill";
Date: Thu, Jul 19, 2018 06:39 PM
To: "Slurm User Community";
Subject: [slurm-users]
Hello,
I've run into a problem passing MCA parameters to openmpi2. This runs
fine on the command-line:
/usr/local/mpi/openmpi2/bin/mpirun --mca btl_tcp_if_include \
192.168.0.0/24 -np 10 -hostfile ~/ompi.hosts \
~/Software/Gulp/gulp-5.0/gulp.ompi example2
If I put the the MCA parameters in ~/op
On Thu, 19 Jul 2018 18:57:09 +0800
"Bill" wrote:
> Hi ,
>
>
> I just found the way , set "DefMemPerCPU=4096" for partition in
> slurm.conf
>
> It will use 4G memory request.
That is how we do it too (except not for a specific partition but
globally).
You can also add custom logic to a submi
Thank you Peter,
Bill
-- Original --
From: Peter Kjellström
Date: Thu,Jul 19,2018 9:51 PM
To: Bill
Cc: Slurm User Community List
Subject: Re: [slurm-users] default memory request
On Thu, 19 Jul 2018 18:57:09 +0800
"Bill" wrote:
> Hi ,
>
>
> I just found t
So the recommendation I've gotten the past is to us option number 4 from
this FAQ:
https://www.open-mpi.org/faq/?category=tuning#setting-mca-params
This works for both mpirun and srun in slurm because its a flat file
that is read rather than options that are passed in.
-Paul Edmon-
On 07/1
Hello Paul,
Paul Edmon writes:
> So the recommendation I've gotten the past is to us option number 4
> from this FAQ:
>
> https://www.open-mpi.org/faq/?category=tuning#setting-mca-params
>
> This works for both mpirun and srun in slurm because its a flat file
> that is read rather than options t
Interesting. Haven't hit that one before. The file has always worked
for me.
-Paul Edmon-
On 07/19/2018 10:28 AM, Roger Mason wrote:
Hello Paul,
Paul Edmon writes:
So the recommendation I've gotten the past is to us option number 4
from this FAQ:
https://www.open-mpi.org/faq/?category=
Roger,
is the home directory mounted at the same place regardless this is a
frontend or a compute node ?
I noted you --export=ALL, so there is a risk you export $HOME that is
not reachable on the compute nodes.
Cheers,
Gilles
- Original Message -
> Hello Paul,
>
> Paul Edmon wri
Hell Gilles,
gil...@rist.or.jp writes:
> is the home directory mounted at the same place regardless this is a
> frontend or a compute node ?
One host serves as both a frontend and compute node and is used to pixie
boot the other compute nodes. On the frontend machine (192.168.0.100) I
have:
mo
Hi all,
We currently have a small Slurm cluster for a research group here, where as a
part of that setup, we have a Slurm DBD setup that we utilize for fair-share
scheduling. The current server platform this runs on is an older Dell PowerEdge
R210 system, with a single 4-core Intel Xeon X3430 C
did you copy the mca parameters file to all the compute nodes as well?
On Thu, Jul 19, 2018 at 11:37 AM, Roger Mason wrote:
> Hell Gilles,
>
> gil...@rist.or.jp writes:
>
>> is the home directory mounted at the same place regardless this is a
>> frontend or a compute node ?
>
> One host serves as
Hello,
Michael Di Domenico writes:
> did you copy the mca parameters file to all the compute nodes as well?
>
No need: my home directory is shared between the submit machine & the
nodes.
Cheers,
Roger
We are pleased to announce the availability of Slurm version 17.11.8.
This includes over 30 fixes made since 17.11.7 was released at the end
of May. This includes a change to the slurmd.service file used with
systemd, this fix prevents systemd from destroying the cgroup
hierarchies slurmd/slur
15 matches
Mail list logo