Failure for 1 node experiments
Hello @froura @agayayav @larriola,
Autosubmit Version
4.1.10
Expid affected(If applicable)
a1p1
Summary
I encountered a problem running a 1 node job. The expid is a1p1. If I set NODES=1 for SIM, it adds the sbatch flag -n 1 (i.e., setting the total number of tasks to 1). Therefore I am getting this error:
+ hres -p 1 -t 4 -h 1 -j 5099589 -J a1p1_19900101_fc0_1_SIM -d 1990010100 -e hz9m -L LWDA -T tco -r 79 -l 137 -f d365 -x '' -N 32 -H mn5-gpp -n 1 -C intel --ifs-bundle-build-dir=/gpfs/projects/ehpc01/models/ifs-nemo/dev-marc/feature-67years/make/marenostrum5-intel/build --icmcl ICMCL_79_1990_extra -R --nemo --nemo-ver=V40 --nemo-grid=eORCA1_Z75 --nemo-xproc=-1 --nemo-yproc=-1 --deep --nonemopart --keepnetcdf --nextgemsout=6 --wam-multio --ifs-multio --restartdirectory=/gpfs/scratch/ehpc01/bsc032655/a1p1/restarts/fc0/current --realization=1 --keeprestart --experiment=control --no-ozone --sicoupl=2 --keepfdb --multio-production-fdbs=/gpfs/scratch/ehpc01/experiments/a1p1/fdb/native:/gpfs/scratch/ehpc01/experiments/a1p1/fdb:/gpfs/scratch/ehpc01/experiments/a1p1/fdb/latlon --outexp=a1p1 --io-tasks=1 --nemo-multio-server-num=1
hres: (Error) Number of compute tasks < 1 -- maybe because # of I/O-nodes (= 0) or I/O-tasks (= 1) was too high ? Yet total # of MPI-tasks = 1 (# of nodes = 1)
The flag is not added for NODES: 2 under SIM.
One can check using autosubmit inspect. In the first case (NODES = 1), I get:
###############################################################################
# SIM a1p1 EXPERIMENT
###############################################################################
#
#SBATCH --qos=gp_ehpc
#SBATCH --partition=standard
#SBATCH --exclusive
#SBATCH -A ehpc01
#
#SBATCH --cpus-per-task=4
#SBATCH --ntasks-per-node=28
#SBATCH --nodes=1
#SBATCH -n 1 <<== THAT'S THE PROBLEM
#
#SBATCH -t 03:00:00
#SBATCH -J a1p1_19900101_fc0_1_SIM
#SBATCH --output=/gpfs/scratch/ehpc01/bsc032655/a1p1/LOG_a1p1/a1p1_19900101_fc0_1_SIM.cmd.out
#SBATCH --error=/gpfs/scratch/ehpc01/bsc032655/a1p1/LOG_a1p1/a1p1_19900101_fc0_1_SIM.cmd.err
#SBATCH --export=ALL
#SBATCH --hint=nomultithread
#
#
###############################################################################
and for the second case (NODES = 2) it does not add the -n (which is correct):
###############################################################################
# SIM a1p1 EXPERIMENT
###############################################################################
#
#SBATCH --qos=gp_ehpc
#SBATCH --partition=standard
#SBATCH --exclusive
#SBATCH -A ehpc01
#
#SBATCH --cpus-per-task=4
#SBATCH --ntasks-per-node=28
#SBATCH --nodes=2
#
#
#SBATCH -t 03:00:00
#SBATCH -J a1p1_19900101_fc0_1_SIM
#SBATCH --output=/gpfs/scratch/ehpc01/bsc032655/a1p1/LOG_a1p1/a1p1_19900101_fc0_1_SIM.cmd.out
#SBATCH --error=/gpfs/scratch/ehpc01/bsc032655/a1p1/LOG_a1p1/a1p1_19900101_fc0_1_SIM.cmd.err
#SBATCH --export=ALL
#SBATCH --hint=nomultithread
#
#
###############################################################################