Changeset 5178 for BOL/LMDZ_Setup_amaury

Timestamp:

Sep 10, 2024, 11:41:21 AM (9 days ago)

Author:

abarral

Message:

Fix libfabric crash on ADS
Increase rebuild job times

Location:

BOL/LMDZ_Setup_amaury

Files:

• lmdz_env.sh (modified) (1 diff)
• reb.sh (modified) (4 diffs)

BOL/LMDZ_Setup_amaury/lmdz_env.sh

@@ -75 +75 @@
       module load PrgEnv-gnu  # we need to load the env because lmdz links some shared libraries
       module load gcc/13.2.0  # required, see https://dci.dci-gitlab.cines.fr/webextranet/user_support/index.html#prgenv-and-compilers
+      export CRAY_CPU_TARGET=x86-64  # to suppress warnings during Cmake netcdf95 build
+      export FI_CXI_RX_MATCH_MODE=hybrid  # 09/24 otherwise we get random SIGABRT e.g. "libfabric:2490616:1725895288::cxi:core:cxip_ux_onload_cb():2657<warn> c1456: RXC (0x5130:21) PtlTE 84:[Fatal] LE resources not recovered during flow control. FI_CXI_RX_MATCH_MODE=[hybrid|software] is required"
 
       function cdo {  # cdo is available as a spack cmd which requires a specific, incompatible env

BOL/LMDZ_Setup_amaury/reb.sh

@@ -21 +21 @@
   job=$SIM$type
 
-  cat <<eod >| $job
+  cat <<eod >| "$job"
 #!/bin/bash
 ## Headers managed by sed
@@ -28 +28 @@
 #@JZ#SBATCH --nodes=1                   # nombre de noeuds
 #@JZ#SBATCH --ntasks-per-node=1         # nombre de taches MPI par noeud
-#@JZ#SBATCH --time=00:30:00             # temps d execution maximum demande (HH:MM:SS)
+#@JZ#SBATCH --time=00:59:00             # temps d execution maximum demande (HH:MM:SS)
 #@JZ#SBATCH --output=post${type}%j.out  # nom du fichier de sortie
 #@JZ#SBATCH --error=post${type}%j.out   # nom du fichier d'erreur (ici en commun avec la sortie)
@@ -35 +35 @@
 #@SP#SBATCH --nodes=1
 #@SP#SBATCH --ntasks-per-node=1
-#@SP#SBATCH --time=00:30:00
+#@SP#SBATCH --time=00:59:00
 #@SP#SBATCH --output=post${type}%j.out
 #@SP#SBATCH --error=post${type}%j.out
@@ -42 +42 @@
 #@ADS#SBATCH --nodes=1
 #@ADS#SBATCH --ntasks-per-node=1
-#@ADS#SBATCH --time=00:30:00
+#@ADS#SBATCH --time=00:59:00
 #@ADS#SBATCH --output=post${type}%j.out
 #@ADS#SBATCH --error=post${type}%j.out