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 <[email protected]> på vegne av Markus Blatt <[email protected]> Sendt: onsdag 11. mars 2020 11:08 Til: [email protected] <[email protected]> Emne: Re: [Opm] OPM Flow multi-node simulations stuck at domain decomposition step 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_CLASS=105 > -x UCX_IB_GID_INDEX=3 --cpu-set > 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35 > -np 144 --hostfile /etc/opt/rdma/hostfile > /mnt/nfs-share/etc/opm-flow/opm-simulators/build/bin/flow > --parameter-file=/mnt/nfs-share/data/norne/params > Out of curiosity. Is there a special reason why --cpu-set runs until 35 with 32 cpus per node? Might you be oversubscribing a node? Markus -- Dr. Markus Blatt OPM-OP AS _______________________________________________ Opm mailing list [email protected] https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fopm-project.org%2Fcgi-bin%2Fmailman%2Flistinfo%2Fopm&data=02%7C01%7CAtgeirr.Rasmussen%40sintef.no%7C92c439a2c2934fdb4f6308d7c5a44bcd%7Ce1f00f39604145b0b309e0210d8b32af%7C1%7C0%7C637195181746088987&sdata=DP6PfdO7olXefHaRUJk6jRg6%2Bmgthhr4lgZTV%2BJ7zIs%3D&reserved=0 _______________________________________________ Opm mailing list [email protected] https://opm-project.org/cgi-bin/mailman/listinfo/opm
