Changeset 3403 for trunk/LMDZ.COMMON/libf/evolution/deftank

Timestamp:

Aug 9, 2024, 6:09:15 PM (5 months ago)

Author:

jbclement

Message:

PEM:
Addition in the launching script of the possibility to submit a job with PBS/TORQUE + Modification to make the time limit detection in "pem.F90" work with PBS/TORQUE + Update of the headers of .job files.
JBC

Location:

trunk/LMDZ.COMMON/libf/evolution/deftank

Files:

• PCMrun.job (modified) (3 diffs)
• PEMrun.job (modified) (2 diffs)
• README (modified) (1 diff)
• lib_launchPEM.sh (modified) (7 diffs)

trunk/LMDZ.COMMON/libf/evolution/deftank/PCMrun.job

@@ -10 +10 @@
 #SBATCH --threads-per-core=1 # --hint=nomultithread
 ###SBATCH --exclusive
-#SBATCH --output=jobPCM_%A.out
+#SBATCH --output=jobPCM_%j.out
 #SBATCH --time=12:00:00
 
@@ -17 +17 @@
 source ../trunk/LMDZ.COMMON/arch.env
 
-# Number of threads to use (must be the same as '#MSUB -c' above)
-export OMP_NUM_THREADS=4
+# Number of threads to use (must be the same as "#SBATCH --cpus-per-task=" above)
+export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
 export OMP_STACKSIZE=400M
 
@@ -24 +24 @@
 exePCM="gcm_64x48x32_phymars_para.e"
 ########################################################################
+
+
+ulimit -s unlimited
 
 # Running the PCM

trunk/LMDZ.COMMON/libf/evolution/deftank/PEMrun.job

@@ -5 +5 @@
 #SBATCH --constraint=GENOA
 ### Number of Nodes to use
-#SBATCH --nodes=4 # to run with enough memory
+#SBATCH --nodes=1
 #SBATCH --ntasks-per-node=1
 #SBATCH --cpus-per-task=1
-#SBATCH --threads-per-core=1 # --hint=nomultithread
 ###SBATCH --exclusive
-#SBATCH --output=jobPEM_%A.out
+#SBATCH --output=jobPEM_%j.out
 #SBATCH --time=24:00:00
 
@@ -24 +23 @@
 ########################################################################
 
+
+ulimit -s unlimited
 
 # Reshaping PCM data with XIOS

trunk/LMDZ.COMMON/libf/evolution/deftank/README

@@ -17 +17 @@
 
 # PCMrun.job:
-  Bash script file to run the PCM with a SLURM job. The name of the PCM executable file should be adapted. The header corresponds to the ADASTRA supercomputer and should be changed for other machines. In case of 1D, the SBATCH headers are omitted.
+  Bash script file to submit a PCM job (with SLURM or PBS/TORQUE). The name of the PCM executable file should be adapted. The headers correspond to the ADASTRA supercomputer and should be changed for other machines and job schedulers. In case of 1D, the headers are naturally omitted.
   The path to source the arch file should be adapted to the machine.
   The execution line should also be adapted according to the set-up.
 
 # PEMrun.job:
-  Bash script file to run the PEM with a SLURM job. The name of the PEM executable file and Reshaping executable file should be adapted. The header corresponds to the ADASTRA supercomputer and should be changed for other machines. In case of 1D, the SBATCH headers are omitted.
+  Bash script file to submit PEM job (with SLURM or PBS/TORQUE). The name of the PEM executable file and Reshaping executable file should be adapted. The headers correspond to the ADASTRA supercomputer and should be changed for other machines and job schedulers. In case of 1D, the headers are naturally omitted.
   The path to source the arch file should be adapted to the machine.
   The execution line should also be adapted according to the set-up.

trunk/LMDZ.COMMON/libf/evolution/deftank/lib_launchPEM.sh

@@ -24 +24 @@
 }
 
-# To check if SLURM is the job scheduler
-function is_slurm() {
-    if ! command -v squeue &> /dev/null; then
-        echo "Error: the job scheduler is not SLURM on $machine!"
-        echo "You need to adapt the script to your case."
+# To check what is the job scheduler
+function job_scheduler() {
+    if command -v squeue &> /dev/null; then
+        echo "SLURM is installed on $machine."
+        name_job="#SBATCH --job-name="
+        kill_job="scancel"
+        submit_job="sbatch --parsable"
+        submit_dependjob="sbatch --parsable --dependency"
+        sed -i 's/\$PBS_JOBID/\$SLURM_JOB_ID/g' PEMrun.job
+    elif command -v qstat &> /dev/null; then
+        echo "PBS/TORQUE is installed on $machine."
+        name_job="#PBS -N "
+        kill_job="qdel"
+        submit_job="qsub"
+        submit_dependjob="qsub -W depend"
+        sed -i 's/\$SLURM_JOB_ID/\$PBS_JOBID/g' PEMrun.job
+    else
+        echo "Error: neither SLURM nor TORQUE/PBS is installed on $machine!"
+        echo "You need to adapt the script to your job scheduler."
         errlaunch
     fi
@@ -39 +53 @@
     LC_NUMERIC=en_US.UTF-8
 
-    if [ $dim -ne 1 ]; then # 3D model works only with SLURM for now (because of chained job submission in the script + time limit management)
-        is_slurm
-    fi
     if [ -v n_mars_years ] && [ ! -z "$n_mars_years" ]; then
         if [ $n_mars_years -lt 1 ]; then
@@ -107 +118 @@
     if [ ! -d "diags" ]; then
         mkdir diags
+    fi
+    if [ $dim -ne 1 ]; then
+        job_scheduler
     fi
 }
@@ -160 +174 @@
         else # 3D model
             cp PCMrun.job PCMrun${iPCM}.job
-            sed -i -E "s/(#SBATCH --job-name=[^0-9]*[0-9]*[^0-9]*)[0-9]+$/\1${iPCM}/" PCMrun${iPCM}.job
+            sed -i -E "s/($name_job[^0-9]*[0-9]*[^0-9]*)[0-9]+$/\1${iPCM}/" PCMrun${iPCM}.job
             sed -i "s/^k=[0-9]\+$/k=$(echo "3 - $nPCM_ini" | bc -l)/" PCMrun${iPCM}.job
-            jobID=$(sbatch --parsable PCMrun${iPCM}.job)
+            jobID=$(eval "$submit_job PCMrun${iPCM}.job")
             # Create a file to cancel the dependent jobs of the cycle
             echo "#!/bin/bash" > kill_launchPEM.sh
             chmod +x kill_launchPEM.sh
-            echo "scancel" $jobID >> kill_launchPEM.sh
+            echo $kill_job $jobID >> kill_launchPEM.sh
         fi
         ((iPCM++))
@@ -182 +196 @@
             else # 3D model
                 cp PCMrun.job PCMrun${iPCM}.job
-                sed -i -E "s/(#SBATCH --job-name=[^0-9]*[0-9]*[^0-9]*)[0-9]+$/\1${iPCM}/" PCMrun${iPCM}.job
+                sed -i -E "s/($name_job[^0-9]*[0-9]*[^0-9]*)[0-9]+$/\1${iPCM}/" PCMrun${iPCM}.job
                 sed -i "s/^k=[0-9]\+$/k=$(echo "$i + 2 - $nPCM_ini" | bc -l)/" PCMrun${iPCM}.job
-                jobID=$(sbatch --parsable --dependency=afterok:${jobID} PCMrun${iPCM}.job)
-                echo "scancel" $jobID >> kill_launchPEM.sh
+                jobID=$(eval "$submit_dependjob=afterok:${jobID} PCMrun${iPCM}.job")
+                echo $kill_job $jobID >> kill_launchPEM.sh
             fi
             ((iPCM++))
@@ -203 +217 @@
             ./PEMrun.job
         else # 3D model
-            sed -i -E "s/(#SBATCH --job-name=[^0-9]*[0-9]*[^0-9]*)[0-9]+$/\1${iPEM}/" PEMrun.job
-            jobID=$(sbatch --parsable PEMrun.job)
+            sed -i -E "s/($name_job[^0-9]*[0-9]*[^0-9]*)[0-9]+$/\1${iPEM}/" PEMrun.job
+            jobID=$(eval "$submit_job PEMrun.job")
             # Create a file to cancel the dependent jobs of the cycle
             echo "#!/bin/bash" > kill_launchPEM.sh
             chmod +x kill_launchPEM.sh
-            echo "scancel" $jobID >> kill_launchPEM.sh
+            echo $kill_job $jobID >> kill_launchPEM.sh
         fi
     else
@@ -229 +243 @@
             ./PEMrun.job
         else # 3D model
-            sed -i -E "s/(#SBATCH --job-name=[^0-9]*[0-9]*[^0-9]*)[0-9]+$/\1${iPEM}/" PEMrun.job
-            jobID=$(sbatch --parsable --dependency=afterok:${jobID} PEMrun.job)
-            echo "scancel" $jobID >> kill_launchPEM.sh
+            sed -i -E "s/($name_job[^0-9]*[0-9]*[^0-9]*)[0-9]+$/\1${iPEM}/" PEMrun.job
+            jobID=$(eval "$submit_dependjob=afterok:${jobID} PEMrun.job")
+            echo $kill_job $jobID >> kill_launchPEM.sh
         fi
     else