Merge branch 'ufs-community:develop' into develop

.cicd/scripts/disk_usage.sh

@@ -0,0 +1,48 @@
+#!/usr/bin/env bash
+
+# Output a CSV report of disk usage on subdirs of some path
+# Usage: 
+#    [JOB_NAME=<ci_job>] [BUILD_NUMBER=<n>] [SRW_COMPILER=<intel>] [SRW_PLATFORM=<machine>] disk_usage path depth size outfile.csv
+#
+# args:
+#    directory=$1
+#    depth=$2
+#    size=$3
+#    outfile=$4
+
+[[ -n ${WORKSPACE} ]] || WORKSPACE=$(pwd)
+[[ -n ${SRW_PLATFORM} ]] || SRW_PLATFORM=$(hostname -s 2>/dev/null) || SRW_PLATFORM=$(hostname 2>/dev/null)
+[[ -n ${SRW_COMPILER} ]] || SRW_COMPILER=compiler
+
+script_dir="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" > /dev/null 2>&1 && pwd)"
+
+# Get repository root from Jenkins WORKSPACE variable if set, otherwise, set
+# relative to script directory.
+declare workspace
+if [[ -n "${WORKSPACE}/${SRW_PLATFORM}" ]]; then
+    workspace="${WORKSPACE}/${SRW_PLATFORM}"
+else
+    workspace="$(cd -- "${script_dir}/../.." && pwd)"
+fi
+
+echo "STAGE_NAME=${STAGE_NAME}" # from pipeline
+outfile="${4:-${workspace}-${SRW_COMPILER}-disk-usage${STAGE_NAME}.csv}"
+
+function disk_usage() {
+    local directory=${1:-${PWD}}
+    local depth=${2:-1}
+    local size=${3:-k}
+    echo "Disk usage: ${JOB_NAME:-ci}/${SRW_PLATFORM}/$(basename $directory)"
+    (
+    cd $directory || exit 1
+    echo "Platform,Build,Owner,Group,Inodes,${size:-k}bytes,Access Time,Filename"
+    du -Px -d ${depth:-1} --inode --exclude='./workspace' | \
+        while read line ; do
+            arr=($line); inode=${arr[0]}; filename=${arr[1]};
+            echo "${SRW_PLATFORM}-${SRW_COMPILER:-compiler},${JOB_NAME:-ci}/${BUILD_NUMBER:-0},$(stat -c '%U,%G' $filename),${inode:-0},$(du -Px -s -${size:-k} --time $filename)" | tr '\t' ',' ;
+        done | sort -t, -k5 -n #-r
+    )
+    echo ""
+}
+
+disk_usage $1 $2 $3 | tee ${outfile}

.cicd/scripts/sbatch_srw_ftest.sh

@@ -7,7 +7,7 @@
 #SBATCH --account=${SRW_PROJECT}
 #SBATCH --qos=batch
 #SBATCH --nodes=1
-#SBATCH --tasks-per-node=24
+#SBATCH --tasks-per-node=12
 #SBATCH --cpus-per-task=1
 #SBATCH -t 00:30:00
 #SBATCH -o log_wrap.%j.log

.cicd/scripts/srw_build.sh

@@ -27,7 +27,8 @@ fi
 # Build and install
 cd ${workspace}/tests
 set +e
-./build.sh ${platform} ${SRW_COMPILER}
+/usr/bin/time -p -f '{\n  "cpu": "%P"\n, "memMax": "%M"\n, "mem": {"text": "%X", "data": "%D", "swaps": "%W", "context": "%c", "waits": "%w"}\n, "pagefaults": {"major": "%F", "minor": "%R"}\n, "filesystem": {"inputs": "%I", "outputs": "%O"}\n, "time": {"real": "%e", "user": "%U", "sys": "%S"}\n}' -o ${WORKSPACE}/${SRW_PLATFORM}-${SRW_COMPILER}-time-srw_build.json \
+    ./build.sh ${platform} ${SRW_COMPILER}
 build_exit=$?
 set -e
 cd -

.cicd/scripts/srw_ftest.sh

@@ -46,7 +46,6 @@ fi
 # Test directories
 we2e_experiment_base_dir="${workspace}/expt_dirs"
 we2e_test_dir="${workspace}/tests/WE2E"
-nco_dir="${workspace}/nco_dirs"
 
 pwd
 
@@ -78,19 +77,17 @@ sed "s|^task_get_extrn_lbcs:|task_get_extrn_lbcs:\n  EXTRN_MDL_SOURCE_BASEDIR_LB
 # Use staged data for HPSS supported machines
 sed 's|^platform:|platform:\n  EXTRN_MDL_DATA_STORES: disk|g' -i ush/config.yaml
 
+# Set OMP_NUM_THREADS_RUN_FCST to 1 in config.yaml
+sed 's|^task_run_fcst:|task_run_fcst:\n  OMP_NUM_THREADS_RUN_FCST: 1|1' -i ush/config.yaml
+
 # Activate the workflow environment ...
 source etc/lmod-setup.sh ${platform,,}
 module use modulefiles
 module load build_${platform,,}_${SRW_COMPILER}
 module load wflow_${platform,,}
 
 [[ ${FORGIVE_CONDA} == true ]] && set +e +u    # Some platforms have incomplete python3 or conda support, but wouldn't necessarily block workflow tests
-# Gaea-C5 special case missing jinja2
-if [ "${platform}" == "gaea-c5" ]; then
-  conda activate workflow_tools
-else
-  conda activate srw_app
-fi
+conda activate srw_app
 set -e -u
 
 # Adjust for strict limitation of stack size 
@@ -111,7 +108,7 @@ cp ${workspace}/ush/wrappers/*.sh .
 export JOBSdir=${workspace}/jobs
 export USHdir=${workspace}/ush
 export OMP_NUM_THREADS=1
-export nprocs=24
+export nprocs=12
 
 [[ -n ${TASKS} ]] || TASKS=(
                 run_make_grid

.cicd/scripts/srw_init.sh

@@ -0,0 +1,38 @@
+#!/usr/bin/env bash
+#
+# A unified init script for the SRW application. This script is expected to
+# fetch initial source for the SRW application for all supported platforms.
+#
+set -e -u -x
+
+script_dir="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" > /dev/null 2>&1 && pwd)"
+
+# Get repository root from Jenkins WORKSPACE variable if set, otherwise, set
+# relative to script directory.
+declare workspace
+if [[ -n "${WORKSPACE}/${SRW_PLATFORM}" ]]; then
+    workspace="${WORKSPACE}/${SRW_PLATFORM}"
+else
+    workspace="$(cd -- "${script_dir}/../.." && pwd)"
+fi
+
+# Normalize Parallel Works cluster platform value.
+declare platform
+if [[ "${SRW_PLATFORM}" =~ ^(az|g|p)clusternoaa ]]; then
+    platform='noaacloud'
+else
+    platform="${SRW_PLATFORM}"
+fi
+
+# Build and install
+cd ${workspace}
+set +e
+/usr/bin/time -p -f '{\n  "cpu": "%P"\n, "memMax": "%M"\n, "mem": {"text": "%X", "data": "%D", "swaps": "%W", "context": "%c", "waits": "%w"}\n, "pagefaults": {"major": "%F", "minor": "%R"}\n, "filesystem": {"inputs": "%I", "outputs": "%O"}\n, "time": {"real": "%e", "user": "%U", "sys": "%S"}\n}' -o ${WORKSPACE}/${SRW_PLATFORM}-${SRW_COMPILER}-time-srw_init.json \
+    ./manage_externals/checkout_externals
+init_exit=$?
+echo "STAGE_NAME=${STAGE_NAME}"
+env | grep = | sort > ${WORKSPACE}/${SRW_PLATFORM}-${SRW_COMPILER}-env.txt
+set -e
+cd -
+
+exit $init_exit

.cicd/scripts/srw_metric.sh

@@ -0,0 +1,146 @@
+#!/usr/bin/env bash
+#
+# The goal of this script is to provide an example of performing Indy-Severe-Weather test run and compare results to reference with
+# Skill score index that is calculated by MET Stat-Analysis Tools
+#
+# Required (these options are set in the Jenkins env):
+#    WORKSPACE=</full/path/to/ufs-srweather-app>
+#    SRW_PLATFORM=<supported_platform_host>
+#    SRW_COMPILER=<intel|gnu>
+#    SRW_PROJECT=<platform_account>
+#
+# Optional:
+#[[ -n ${SRW_PROJECT} ]] || SRW_PROJECT="no_account"
+[[ -n ${FORGIVE_CONDA} ]] || FORGIVE_CONDA=true
+set -e -u -x
+
+BUILD_OPT=false
+RUN_WE2E_OPT=false
+RUN_STAT_ANLY_OPT=false
+
+if [[ $# -eq 0 ]]; then
+    BUILD_OPT=true
+    RUN_WE2E_OPT=true
+    RUN_STAT_ANLY_OPT=true
+elif [[ $# -ge 4 ]]; then 
+    echo "Too many arguments, expecting three or less"
+    exit 1
+else 
+    for opt in "$@"; do    
+       case $opt in
+            build) BUILD_OPT=true ;; 
+            run_we2e) RUN_WE2E_OPT=true ;;
+            run_stat_anly) RUN_STAT_ANLY_OPT=true ;;
+            *) echo "Not valid option. Exiting!" ; exit 1 ;;
+        esac
+    done
+fi
+
+script_dir="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" > /dev/null 2>&1 && pwd)"
+
+# Get repository root from Jenkins WORKSPACE variable if set, otherwise, set
+# relative to script directory.
+declare workspace
+if [[ -d "${WORKSPACE}/${SRW_PLATFORM}" ]]; then
+    workspace="${WORKSPACE}/${SRW_PLATFORM}"
+else
+    workspace="$(cd -- "${script_dir}/../.." && pwd)"
+fi
+
+# Normalize Parallel Works cluster platform value.
+declare platform
+if [[ "${SRW_PLATFORM}" =~ ^(az|g|p)clusternoaa ]]; then
+    platform='noaacloud'
+else
+    platform="${SRW_PLATFORM}"
+fi
+
+# Test directories
+we2e_experiment_base_dir="${we2e_experiment_base_dir:=${workspace}/../expt_dirs/metric_test}"
+we2e_test_dir="${we2e_test_dir:=${workspace}/tests/WE2E}"
+we2e_test_name="grid_SUBCONUS_Ind_3km_ics_FV3GFS_lbcs_FV3GFS_suite_WoFS_v0"
+
+pwd
+
+# Setup the build environment
+declare srw_compiler
+srw_compiler=${SRW_COMPILER} 
+source ${workspace}/etc/lmod-setup.sh ${platform,,}
+module use ${workspace}/modulefiles
+module load build_${platform,,}_${srw_compiler}
+
+# Build srw
+if [[ ${BUILD_OPT} == true ]]; then
+    cd ${workspace}/tests
+    ./build.sh ${platform,,} ${srw_compiler}
+fi
+cd ${workspace}
+
+# Activate workflow environment
+module load wflow_${platform,,}
+
+[[ ${FORGIVE_CONDA} == true ]] && set +e +u    # Some platforms have incomplete python3 or conda support, but would not necessarily block workflow tests
+conda activate srw_app
+set -e -u
+
+# Run test
+declare srw_project
+srw_project=${SRW_PROJECT}
+if [[ ${RUN_WE2E_OPT} == true ]]; then
+    [[ -d ${we2e_experiment_base_dir} ]] && rm -rf ${we2e_experiment_base_dir}
+    cd ${workspace}/tests/WE2E
+    ./run_WE2E_tests.py -t ${we2e_test_name} -m ${platform,,} -a ${srw_project} --expt_basedir "metric_test" --exec_subdir=install_intel/exec -q
+fi
+cd ${workspace}
+
+# Run skill-score check
+if [[ ${RUN_STAT_ANLY_OPT} == true ]]; then
+    # Clear out data
+    rm -rf ${workspace}/Indy-Severe-Weather/
+    # Check if metprd data exists locally otherwise get it from S3
+    TEST_EXTRN_MDL_SOURCE_BASEDIR=$(grep TEST_EXTRN_MDL_SOURCE_BASEDIR ${workspace}/ush/machine/${SRW_PLATFORM}.yaml | awk '{print $NF}')
+    if [[ -d $(dirname ${TEST_EXTRN_MDL_SOURCE_BASEDIR})/metprd/point_stat ]] ; then
+        mkdir -p Indy-Severe-Weather/metprd/point_stat
+        cp -rp $(dirname ${TEST_EXTRN_MDL_SOURCE_BASEDIR})/metprd/point_stat Indy-Severe-Weather/metprd
+    elif [[ -f Indy-Severe-Weather.tgz ]]; then
+        tar xvfz Indy-Severe-Weather.tgz 
+    else
+        wget https://noaa-ufs-srw-pds.s3.amazonaws.com/sample_cases/release-public-v2.1.0/Indy-Severe-Weather.tgz
+        tar xvfz Indy-Severe-Weather.tgz
+    fi
+    [[ -f ${platform,,}-${srw_compiler}-skill-score.txt ]] && rm ${platform,,}-${srw_compiler}-skill-score.txt
+    # Skill score index is computed over several terms that are defined in parm/metplus/STATAnalysisConfig_skill_score. 
+    # It is computed by aggregating the output from earlier runs of the Point-Stat and/or Grid-Stat tools over one or more cases.
+    # In this example, skill score index is a weighted average of 4 skill scores of RMSE statistics for wind speed, dew point temperature, 
+    # temperature, and pressure at lowest level in the atmosphere over 6 hour lead time.
+    cp ${we2e_experiment_base_dir}/${we2e_test_name}/2019061500/metprd/PointStat/*.stat ${workspace}/Indy-Severe-Weather/metprd/point_stat/
+    # Remove conda for Orion due to conda env conflicts
+    if [[ ${platform} =~ "orion" ]]; then
+        sed -i 's|load("conda")|--load("conda")|g' ${workspace}/modulefiles/tasks/${platform,,}/run_vx.local.lua
+    fi
+    # Load met and metplus
+    module use modulefiles/tasks/${platform,,}
+    module load run_vx.local 
+    # Reset Orion run_vx.local file
+    if [[ ${platform} =~ "orion" ]]; then
+       sed -i 's|--load("conda")|load("conda")|g' ${workspace}/modulefiles/tasks/${platform,,}/run_vx.local.lua
+    fi
+    # Run stat_analysis
+    stat_analysis -config parm/metplus/STATAnalysisConfig_skill_score -lookin ${workspace}/Indy-Severe-Weather/metprd/point_stat -v 2 -out ${platform,,}-${srw_compiler}-skill-score.txt
+
+    # check skill-score.txt
+    cat ${platform,,}-${srw_compiler}-skill-score.txt
+
+    # get skill-score (SS_INDEX) and check if it is significantly smaller than 1.0
+    # A value greater than 1.0 indicates that the forecast model outperforms the reference, 
+    # while a value less than 1.0 indicates that the reference outperforms the forecast.
+    tmp_string=$( tail -2 ${platform,,}-${srw_compiler}-skill-score.txt | head -1 )
+    SS_INDEX=$(echo $tmp_string | awk -F " " '{print $NF}')
+    echo "Skill Score: ${SS_INDEX}"
+    if [[ ${SS_INDEX} < "0.700" ]]; then
+        echo "Your Skill Score is way smaller than 1.00, better check before merging"
+        exit 1
+    else
+        echo "Congrats! You pass check!"
+   fi
+fi