Hi, Yogi,
I see you are trying to run Norne with 144 processes. Did you see the problem
with much less processes, for example 4 or 8 processes?
In my opinion, with the current approach for domain decomposition, it can be
challenging to run Norne with so many processes, considering the relative
Hello everyone,
I am trying to run steadystate_test_upscaling (from opm-upscaling) on a
primary drainage case, and I am struggling to get convincing results.
No matter what capillary number I use, the output relperms stay the
same (down to the last digit).
We are suspecting it might have to do wit
On Tue, Mar 10, 2020 at 02:15:19PM -0700, Yogi Pandey wrote:
> All,
> I am trying to run OPM Flow simulations on multiple nodes. I have built OPM
> Flow from source on Oracle Linux 7 OS (binary compatible with RHEL) with:
>
> [...]
>
> .OPM Flow modules are built using following commads:
Hi Yogi,
On Tue, Mar 10, 2020 at 02:15:19PM -0700, Yogi Pandey wrote:
> Simulation is being run on 4 nodes with 32 processors each using following
> command:
>
> mpirun --display-map -mca btl self -x UCX_TLS=rc,self,sm -x
> HCOLL_ENABLE_MCAST_ALL=0 -mca coll_hcoll_enable 0 -x UCX_IB_TRAFFIC_CLA
Hi Yogi,
Lots of things have changed in Flow for the initialization in parallel
recently. Could you try to check out the previous release (2019.10) and check
if you see the same problems?
Atgeirr
Frå: Opm på vegne av Markus Blatt
Sendt: onsdag 11. mars 2020 1
Dear opm community;
we are approaching April and the 2020.04 release of opm/flow is soon
imminent, and it is time to start planning the release process. If someone
has planned to be the release manager for the 2020.04 release that is cool,
otherwise opm-op can be responsible for managing the 2020.
Hi Kai,
Thank you for your reply. I have run the Norne model with 1, 2, 4, 8, 16, 36,
72, and 108 processes successfully. The run gets stuck at 144 processes. As I
mentioned, I did try a larger model, but even there I was not able to scale on
more than 1 node for a certain configuration. Fin
Hi Dr. Blatt,
I am using OPM Flow version 2020.04-pre. Each node has 384 GB memory, and I
don't see any memory problem indicated by kswapd appearing.
With some hit and trial I arrived at a model with just 1 producer and 1
injector, with 11 million cells, which scales on 4 nodes. My impression i
Dr. Blatt,
That's a typo. Each node has 36 processors.
Thank you,
Yogi
-Original Message-
From: Markus Blatt [mailto:mar...@dr-blatt.de]
Sent: Wednesday, March 11, 2020 5:09 AM
To: opm@opm-project.org
Subject: Re: [Opm] OPM Flow multi-node simulations stuck at domain
decomposition step
Thank you for the suggestion Atgeirr. I'll try out the previous release 2019.10.
- Yogi
-Original Message-
From: Atgeirr Rasmussen [mailto:atgeirr.rasmus...@sintef.no]
Sent: Wednesday, March 11, 2020 6:46 AM
To: opm@opm-project.org
Subject: Re: [Opm] OPM Flow multi-node simulations stuck
10 matches
Mail list logo