[slurm-users] SLURM A100

Kota Tsuyuzaki kota.tsuyuzaki.pc at hco.ntt.co.jp
Wed Apr 21 09:09:28 UTC 2021

Hello Tim,

In the last year, I figured out the A100 MIG feature behavior with Slurm Workload Manager. At that time, it required non-default
DEVFS mode in kernel config to constraint the MIG device via Slurm cgroup. After the setting, A100 MIG works well to me so I suppose
it should NOT be blocking issue except you need to have the configuration.

My testing NVIDIA driver version was at 450.51.06, and the mode was not default at that time but the NVIDIA documents said the DEVFS
mode will be default in the future so that you should check the current newest docs if you mind the kernel setting.

The procedure how we can configure the DEVFS mode to A100 was written to my blog post(*1). It's so sorry that was in Japanese but
hopefully, the setting scripts and web links to NVIDIA official documents would be helpful for you. Perhaps, google translation too.

1: https://medium.com/nttlabs/nvidia-a100-mig-as-linux-device-66220ca16698


露崎 浩太 (Kota Tsuyuzaki)
kota.tsuyuzaki.pc at hco.ntt.co.jp

> -----Original Message-----
> From: slurm-users <slurm-users-bounces at lists.schedmd.com> On Behalf Of Timothy Carr
> Sent: Wednesday, April 21, 2021 4:14 PM
> To: slurm-users at schedmd.com
> Subject: [slurm-users] SLURM A100
> Dear Community,
> Trust everyone is well and keeping safe?
> We are considering the purchase of nodes with the Nvidia A100 GPUs and enabling the MIG feature which allows for the
> creation of instance resource profiles. The creation of these profiles seems to be straightforward as per the
> documentation. Have any of you had the opportunity to implement the A100 MIG with SLURM and have you found any
> caveats you are willing to share?
> Kind Regards
> --
> Tim
> Disclaimer - University of Cape Town This email is subject to UCT policies and email disclaimer published on our website
> at http://www.uct.ac.za/main/email-disclaimer or obtainable from +27 21 650 9111. If this email is not related to the
> business of UCT, it is sent by the sender in an individual capacity. Please report security incidents or abuse via
> https://csirt.uct.ac.za/page/report-an-incident.php.

More information about the slurm-users mailing list