Work update 20250117

tests/newtests/detect_machine.sh

@@ -0,0 +1,110 @@
+#!/bin/bash
+
+# The authoritative copy of this script lives in the ufs-weather-model at:
+# https://github.com/ufs-community/ufs-weather-model/blob/develop/tests/detect_machine.sh
+# If any local modifications are made or new platform support added,
+# please consider opening an issue and a PR to the ufs-weather-model
+# so that this copy remains in sync with its authoritative source
+#
+# Thank you for your contribution
+
+# If the MACHINE_ID variable is set, skip this script.
+[[ -n ${MACHINE_ID:-} ]] && return
+
+# First detect w/ hostname
+case $(hostname -f) in
+
+  adecflow0[12].acorn.wcoss2.ncep.noaa.gov)  MACHINE_ID=acorn ;; ### acorn
+  alogin0[123].acorn.wcoss2.ncep.noaa.gov)    MACHINE_ID=acorn ;; ### acorn
+  clogin0[1-9].cactus.wcoss2.ncep.noaa.gov)  MACHINE_ID=wcoss2 ;; ### cactus01-9
+  clogin10.cactus.wcoss2.ncep.noaa.gov)      MACHINE_ID=wcoss2 ;; ### cactus10
+  dlogin0[1-9].dogwood.wcoss2.ncep.noaa.gov) MACHINE_ID=wcoss2 ;; ### dogwood01-9
+  dlogin10.dogwood.wcoss2.ncep.noaa.gov)     MACHINE_ID=wcoss2 ;; ### dogwood10
+
+  gaea5[1-8])          MACHINE_ID=gaeac5 ;; ### gaea51-58
+  gaea5[1-8].ncrc.gov) MACHINE_ID=gaeac5 ;; ### gaea51-58
+  gaea6[1-8])          MACHINE_ID=gaeac6 ;; ### gaea61-68
+  gaea6[1-8].ncrc.gov) MACHINE_ID=gaeac6 ;; ### gaea61-68
+
+  hfe0[1-9]) MACHINE_ID=hera ;; ### hera01-09
+  hfe1[0-2]) MACHINE_ID=hera ;; ### hera10-12
+  hecflow01) MACHINE_ID=hera ;; ### heraecflow01
+
+  s4-submit.ssec.wisc.edu) MACHINE_ID=s4 ;; ### s4
+
+  fe[1-8]) MACHINE_ID=jet ;; ### jet01-8
+  tfe[12]) MACHINE_ID=jet ;; ### tjet1-2
+
+  Orion-login-[1-4].HPC.MsState.Edu) MACHINE_ID=orion ;; ### orion1-4
+
+  [Hh]ercules-login-[1-4].[Hh][Pp][Cc].[Mm]s[Ss]tate.[Ee]du) MACHINE_ID=hercules ;; ### hercules1-4
+
+  derecho1.hsn.de.hpc.ucar.edu) MACHINE_ID=derecho ;; ### derecho1
+  derecho2.hsn.de.hpc.ucar.edu) MACHINE_ID=derecho ;; ### derecho2
+  derecho3.hsn.de.hpc.ucar.edu) MACHINE_ID=derecho ;; ### derecho3
+  derecho4.hsn.de.hpc.ucar.edu) MACHINE_ID=derecho ;; ### derecho4
+  derecho5.hsn.de.hpc.ucar.edu) MACHINE_ID=derecho ;; ### derecho5
+  derecho6.hsn.de.hpc.ucar.edu) MACHINE_ID=derecho ;; ### derecho6
+  derecho7.hsn.de.hpc.ucar.edu) MACHINE_ID=derecho ;; ### derecho7
+  derecho8.hsn.de.hpc.ucar.edu) MACHINE_ID=derecho ;; ### derecho8
+
+  login[1-4].stampede2.tacc.utexas.edu) MACHINE_ID=stampede ;; ### stampede1-4
+
+  login[1-4].frontera.tacc.utexas.edu) MACHINE_ID=frontera ;; ### frontera1-4
+  c*.frontera.tacc.utexas.edu) MACHINE_ID=frontera ;; ### frontera compute 
+
+  login0[1-2].expanse.sdsc.edu) MACHINE_ID=expanse ;; ### expanse1-2
+
+  discover3[1-5].prv.cube) MACHINE_ID=discover ;; ### discover31-35
+  *) MACHINE_ID=UNKNOWN ;;  # Unknown platform
+esac
+
+if [[ ${MACHINE_ID} == "UNKNOWN" ]]; then 
+   case ${PW_CSP:-} in
+      "aws" | "google" | "azure") MACHINE_ID=noaacloud ;;
+      *) PW_CSP="UNKNOWN"
+   esac
+fi
+
+# Overwrite auto-detect with MACHINE if set
+MACHINE_ID=${MACHINE:-${MACHINE_ID}}
+
+# If MACHINE_ID is no longer UNKNNOWN, return it
+if [[ "${MACHINE_ID}" != "UNKNOWN" ]]; then
+  return
+fi
+
+# Try searching based on paths since hostname may not match on compute nodes
+if [[ -d /lfs/h3 ]]; then
+  # We are on NOAA Cactus or Dogwood
+  MACHINE_ID=wcoss2
+elif [[ -d /lfs/h1 && ! -d /lfs/h3 ]]; then
+  # We are on NOAA TDS Acorn
+  MACHINE_ID=acorn
+elif [[ -d /mnt/lfs1 ]]; then
+  # We are on NOAA Jet
+  MACHINE_ID=jet
+elif [[ -d /scratch1 ]]; then
+  # We are on NOAA Hera
+  MACHINE_ID=hera
+elif [[ -d /work ]]; then
+  # We are on MSU Orion or Hercules
+  mount=$(findmnt -n -o SOURCE /home)   
+  if [[ ${mount} =~ "hercules" ]]; then
+    # We are on Hercules
+    MACHINE_ID=hercules
+  else
+    MACHINE_ID=orion
+  fi
+elif [[ -d /gpfs/f5 && -d /ncrc ]]; then
+  # We are on GAEA C5.
+  MACHINE_ID=gaeac5
+elif [[ -d /gpfs/f6 && -d /ncrc ]]; then
+  # We are on GAEA C6.
+  MACHINE_ID=gaeac6
+elif [[ -d /data/prod ]]; then
+  # We are on SSEC's S4
+  MACHINE_ID=s4
+else
+  echo WARNING: UNKNOWN PLATFORM 1>&2
+fi

tests/newtests/error_handling.sh

@@ -0,0 +1,35 @@
+#! /bin/bash
+
+die() { echo "$@" >&2; exit 1; }
+
+handle_error() {
+  echo "rt.sh: Getting error information..."
+  local exit_code=$1
+  local exit_line=$2
+  echo "Exited at line ${exit_line} having code ${exit_code}"
+  rt_trap
+}
+
+rt_trap() {
+  echo "rt.sh: Exited abnormally, killing workflow and cleaning up"
+  trap "" SIGINT
+  [[ ${ROCOTO:-false} == true ]] && rocoto_kill
+  [[ ${ECFLOW:-false} == true ]] && ecflow_kill
+  cleanup
+}
+
+cleanup() {
+  echo "rt.sh: Cleaning up..."
+  awk_info=$(awk '{print $2}' < "${LOCKDIR}/PID")
+  [[ ${awk_info} == "$$" ]] && rm -rf "${LOCKDIR}"
+  [[ ${ECFLOW:-false} == true ]] && ecflow_stop
+  trap 0
+  echo "rt.sh: Exiting."
+  exit
+}
+
+trap '{ echo "rt.sh interrupted"; rt_trap ; }' INT
+trap '{ echo "rt.sh quit"; rt_trap ; }' QUIT
+trap '{ echo "rt.sh terminated"; rt_trap ; }' TERM
+trap '{ handle_error $? $LINENO ; }' ERR
+trap '{ echo "rt.sh finished"; cleanup ; }' EXIT

tests/newtests/manage_ecflow.sh

@@ -0,0 +1,95 @@
+#! /bin/bash
+
+source error_handling.sh
+
+case ${MACHINE_ID} in
+  wcoss2|acorn)
+    module load ecflow/5.6.0.13
+    ;;
+  gaeac5)
+    # UFSWM Uses ECFLow in Python
+    module load PrgEnv-intel/8.5.0
+    module load intel-classic/2023.2.0
+    module load python/3.9.12
+
+    # ECFLow isn't installed by default. It is in epic stack
+    module use /ncrc/proj/epic/spack-stack/modulefiles
+    module load ecflow/5.8.4
+    ECF_HOST=$(hostname)
+    ECF_PORT=$(( $(id -u) + 1500 ))
+    export ECF_PORT ECF_HOST
+    ;;
+  gaeac6)
+    # UFSWM Uses ECFLow in Python
+    module use /ncrc/proj/epic/spack-stack/c6/spack-stack-1.6.0/envs/fms-2024.01/install/modulefiles/Core
+    module load stack-intel/2023.2.0
+    module load python/3.10.13
+
+    # ECFLow isn't installed by default. It is in epic stack
+    module use /ncrc/proj/epic/spack-stack/modulefiles
+    module load ecflow/5.8.4
+    ECF_HOST=$(hostname)
+    ECF_PORT=$(( $(id -u) + 1500 ))
+    export ECF_PORT ECF_HOST
+    ;;
+  hera)
+    module load ecflow/5.11.4
+    ;;
+  orion)
+    module use /work/noaa/epic/role-epic/spack-stack/orion/modulefiles
+    module load ecflow/5.8.4
+    ECF_HOST=$(hostname)
+    ECF_PORT="$(( $(id -u) + 1500 ))"
+    export ECF_PORT ECF_HOST
+    ;;
+  hercules)
+    module use /work/noaa/epic/role-epic/spack-stack/hercules/modulefiles
+    module load ecflow/5.8.4
+    ECF_HOST=$(hostname)
+    ECF_PORT="$(( $(id -u) + 1500 ))"
+    export ECF_PORT ECF_HOST
+    ;;
+  jet)
+    # UFSWM Uses ECFlow in Python
+    module use /contrib/spack-stack/spack-stack-1.6.0/envs/unified-env-rocky8/install/modulefiles/Core
+    module load stack-intel/2021.5.0
+    module load stack-python/3.10.13
+
+    module load ecflow/5.11.4
+    ;;
+  s4)
+    module load ecflow/5.6.0
+    module load miniconda/3.8-s4
+    module use /data/prod/jedi/spack-stack/modulefiles
+    module load ecflow/5.8.4
+    ECF_HOST=$(hostname)
+    ECF_PORT="$(( $(id -u) + 1500 ))"
+    export ECF_PORT ECF_HOST
+    ;;
+  derecho)
+    # ECFlow needs Python
+    module unload ncarcompilers
+    module use /glade/work/epicufsrt/contrib/spack-stack/derecho/spack-stack-1.5.1/envs/unified-env/install/modulefiles/Core
+    module load stack-intel/2021.10.0
+    module load stack-python/3.10.8
+
+    # ECFlow not in default system stack. Need to use epic build
+    module use /glade/work/epicufsrt/contrib/spack-stack/derecho/modulefiles
+    module load ecflow/5.8.4
+    ECF_HOST=$(hostname)
+    ECF_PORT=$(( $(id -u) + 1500 ))
+    export ECF_PORT ECF_HOST
+    ;;
+  stampede)
+    die "Machine does not support ECFlow."
+    ;;
+  expanse)
+    die "Machine does not support ECFlow."
+    ;;
+  noaacloud)
+    die "Machine does not support ECFlow."
+    ;;
+  *)
+    die "Unknown machine ID, please edit detect_machine.sh file"
+    ;;
+esac

tests/newtests/manage_rocoto.sh

@@ -0,0 +1,59 @@
+#! /bin/bash
+
+source error_handling.sh
+
+case ${MACHINE_ID} in
+  wcoss2|acorn)
+    ROCOTO_SCHEDULER="pbs"
+    ;;
+  gaeac5)
+    module use /ncrc/proj/epic/rocoto/modulefiles
+    module load rocoto
+    ROCOTO_SCHEDULER="slurm"
+    ;;
+  gaeac6)
+    module use /ncrc/proj/epic/c6/modulefiles
+    module load rocoto/1.3.7
+    ROCOTO_SCHEDULER="slurm"
+    ;;
+  hera)
+    module load rocoto
+    ROCOTO_SCHEDULER=slurm
+    ;;
+  orion)
+    module load contrib ruby/3.2.3 rocoto/1.3.7
+    ROCOTO_SCHEDULER="slurm"
+    ;;
+  hercules)
+    module load contrib rocoto
+    ROCOTO_SCHEDULER="slurm"
+    ;;
+  jet)
+    module load rocoto
+    ROCOTO_SCHEDULER="slurm"
+    ;;
+  s4)
+    module load rocoto/1.3.2
+    ROCOTO_SCHEDULER=slurm
+    ;;
+  derecho)
+    module use /glade/work/epicufsrt/contrib/derecho/rocoto/modulefiles
+    module load rocoto
+    ROCOTO_SCHEDULER="pbspro"
+    ;;
+  stampede)
+    die "Machine does not support Rocoto."
+    ;;
+  expanse)
+    die "Machine does not support Rocoto."
+    ;;
+  noaacloud)
+    export PATH="/contrib/EPIC/bin:${PATH}"
+    module use /apps/modules/modulefiles
+    module load rocoto/1.3.7
+    ROCOTO_SCHEDULER=slurm
+    ;;
+  *)
+    die "Unknown machine ID, please edit detect_machine.sh file"
+    ;;
+esac