Adjust C768 resources for Hera (#2819)

This modifies the resources for gdasfcst (everywhere) and enkfgdaseupd (Hera only). For the fcst job, the number of write tasks is increased to prevent out of memory errors from the inline post. For the eupd, the number of tasks is decreased to prevent out of memory errors. The runtime for the eupd job was just over 10 minutes. Resolves #2506 Resolves #2498 Resolves #2916 --------- Co-authored-by: Walter Kolczynski - NOAA <[email protected]>
NOAA-EMC · Sep 20, 2024 · 3c86873 · 3c86873
1 parent c6e3262
commit 3c86873
Show file tree

Hide file tree

Showing 7 changed files with 24 additions and 15 deletions.
diff --git a/jobs/JGFS_ATMOS_CYCLONE_GENESIS b/jobs/JGFS_ATMOS_CYCLONE_GENESIS
@@ -3,11 +3,6 @@
 source "${HOMEgfs}/ush/preamble.sh"
 source "${HOMEgfs}/ush/jjob_header.sh" -e "genesis" -c "base genesis"
 
-# Hack to temporary skip this as the tracker has not been build
-#   on Hercules Rocky 9 yet
-# TODO: Remove this after tracker has been built for Rocky 9 #2639
-if [[ "${machine}" == 'HERCULES' ]]; then exit 0; fi
-
 ##############################################
 # Set variables used in the exglobal script
 ##############################################

diff --git a/jobs/JGFS_ATMOS_CYCLONE_TRACKER b/jobs/JGFS_ATMOS_CYCLONE_TRACKER
@@ -3,11 +3,6 @@
 source "${HOMEgfs}/ush/preamble.sh"
 source "${HOMEgfs}/ush/jjob_header.sh" -e "tracker" -c "base tracker"
 
-# Hack to temporary skip this as the tracker has not been build
-#   on Hercules Rocky 9 yet
-# TODO: Remove this after tracker has been built for Rocky 9 #2639
-if [[ "${machine}" == 'HERCULES' ]]; then exit 0; fi
-
 export COMPONENT="atmos"
 
 

diff --git a/parm/config/gfs/config.base b/parm/config/gfs/config.base
@@ -492,4 +492,11 @@ if [[ "${machine}" =~ "PW" ]]; then
    export DO_WAVE="NO"
 fi
 
+# The tracker and genesis are not installed on Orion/Hercules yet; this requires spack-stack builds of the package.
+# TODO: we should place these in workflow/hosts/[orion|hercules].yaml.
+if [[ "${machine}" == "ORION" || "${machine}" == "HERCULES" ]]; then
+   export DO_TRACKER="NO"
+   export DO_GENESIS="NO"
+fi
+
 echo "END: config.base"
diff --git a/parm/config/gfs/config.resources.HERA b/parm/config/gfs/config.resources.HERA
@@ -5,8 +5,9 @@
 case ${step} in
   "anal")
     if [[ "${CASE}" == "C384" ]]; then
-      export ntasks=270
-      export threads_per_task_anal=8
+      export ntasks_gdas=270
+      export ntasks_gfs=270
+      export threads_per_task=8
       export tasks_per_node=$(( max_tasks_per_node / threads_per_task ))
     fi
     ;;
@@ -26,6 +27,10 @@ case ${step} in
 
   "eupd")
     case ${CASE} in
+      "C768")
+        export ntasks=80
+        export threads_per_task=20
+        ;;
       "C384")
         export ntasks=80
         ;;
@@ -43,6 +48,13 @@ case ${step} in
     export tasks_per_node=$(( max_tasks_per_node / threads_per_task ))
     ;;
 
+  "upp")
+    if (( "${CASE:1}" >= 768 )); then
+       # Run fewer tasks per node for memory
+       tasks_per_node=20
+    fi
+    ;;
+
   *)
     ;;
 esac
diff --git a/parm/config/gfs/config.resources.S4 b/parm/config/gfs/config.resources.S4
@@ -32,7 +32,7 @@ case ${step} in
       *)
         ;;
     esac
-    export tasks_node=$(( max_tasks_per_node / threads_per_task ))
+    export tasks_per_node=$(( max_tasks_per_node / threads_per_task ))
     ;;
 
   "eobs")

diff --git a/parm/config/gfs/config.ufs b/parm/config/gfs/config.ufs
@@ -281,7 +281,7 @@ case "${fv3_res}" in
           export rf_cutoff=100.0
           export fv_sg_adj=450
           export WRITE_GROUP_GDAS=2
-          export WRTTASK_PER_GROUP_PER_THREAD_PER_TILE_GDAS=10
+          export WRTTASK_PER_GROUP_PER_THREAD_PER_TILE_GDAS=15
           export WRITE_GROUP_GFS=4
           export WRTTASK_PER_GROUP_PER_THREAD_PER_TILE_GFS=20 #Note this should be 10 for WCOSS2
         fi

diff --git a/sorc/gsi_utils.fd b/sorc/gsi_utils.fd