Branch / tenant: dev/gfs.v17 (gw_v17) Path: ecf/scripts/gfs/product/wave/gempak/jgfs_wave_gempak_master.ecf Analysis date: 2026-06-11 Method: AWS Bedrock AgentCore agentcore-mcp-rag MCP (graph traversal, GraphRAG, code-context) cross-checked against the on-disk v17 worktree.

jgfs_wave_gempak_master.ecf is the NCO production ecFlow task template that converts GFS wave model gridded output (GRIB2) into GEMPAK format for NAWIPS display and distribution. It is a _master template deployed per forecast hour. At runtime it calls the shared JGLOBAL_WAVE_GEMPAK J-Job, which drives exgfs_wave_nawips.sh to process all configured wave grids for the given forecast hour.

#PBS -S /bin/bash
#PBS -N gfs_wave_gempak_%CYC%_f%FHR%
#PBS -j oe
#PBS -q %QUEUE%
#PBS -A %PROJ%-%PROJENVIR%
#PBS -l walltime=00:20:00
#PBS -l select=1:mpiprocs=128:ompthreads=1:ncpus=128:mem=1GB
#PBS -l place=vscatter:shared
#PBS -l debug=true

model=gfs
export cyc="%CYC%"
%include <head.h>
%include <envir-p1.h>

############################################################
# Load modules
############################################################
source ${HOMEgfs}/dev/ush/load_modules.sh run

module list
export WGF="%WGF%"
export FHR="%FHR%"
export FHR_LIST="%FHR_LIST%"
export FORECAST_HOUR="${FHR_LIST}"
export COMPONENT=${WGF}
############################################################
# CALL executable job script here
############################################################
export CDATE=${PDY}${cyc}
export HOMEglobal=${HOMEgfs}

#### ${HOMEgfs}/dev/job_cards/rocoto/wavegempak.sh
${HOMEgfs}/jobs/JGLOBAL_WAVE_GEMPAK
if [ $? -ne 0 ]; then
   ecflow_client --msg="***JOB ${ECF_NAME} ERROR RUNNING J-SCRIPT ***"
   ecflow_client --abort
   exit
fi

%include <tail.h>
%manual

%end

Single .ecf backs per-forecast-hour ecFlow tasks (jgfs_wave_gempak_f000.ecf, jgfs_wave_gempak_f003.ecf, etc.). Each task overrides FHR and FHR_LIST.

ecf/defs/gfs_prod.def : task jgfs_wave_gempak_fNNN
  +-- jgfs_wave_gempak_master.ecf                (PBS submit; deployed as *_fNNN.ecf)
       +-- ${HOMEgfs}/jobs/JGLOBAL_WAVE_GEMPAK    (J-Job, shared GFS/GDAS)
            +-- wave_domain_grid.sh               (sourced: grid config helper)
            +-- ${SCRglobal}/exgfs_wave_nawips.sh  (ex-script)
                 +-- nagrib2_nc per wave grid      (GRIB2 to GEMPAK conversion)

Confirmed via MCP graph (tenant gw_v17): JGLOBAL_WAVE_GEMPAK is a suite leaf with 0 direct dependents.

The J-Job iterates over all configured wave grids (GEMPAK_GRIDS from config.wavegempak) and converts their GRIB2 output into GEMPAK grid files:

The wave_domain_grid.sh helper is sourced to resolve each grid ID into region/resolution metadata used to construct dynamic COMIN paths (COMIN_WAVE_GRID_${REGION}_${RES}).

DBN alert type: GFS_WAVE_GEMPAK

Uses the EE2-correct pattern: "${SCRglobal}/exgfs_wave_nawips.sh" && true followed by err_exit on non-zero.

The J-Job sources jjob_header.sh -e "wavegempak" -c "base wave wavegempak", loading:

• config.base — fundamental paths.
• config.wave — wave model settings.
• config.wavegempak — wave GEMPAK-specific grid list and settings.

• Shared J-Job: JGLOBAL_WAVE_GEMPAK is used by both GFS and GDAS wave GEMPAK tasks.
• Dynamic grid iteration: uses wave_domain_grid.sh to resolve grid IDs into structured metadata, allowing flexible grid configuration without code changes.
• High CPU / low memory: 128 cpus for parallel nagrib across multiple grids but only 1 GB memory (wave grid conversion is compute-bound, not memory-bound).
• _master template with per-hour deployment.
• Centralized module loading via load_modules.sh run.
• HOMEglobal shim bridges v17 naming to NCO.

The J-Job is a suite leaf. However, wave_domain_grid.sh is shared by many wave processing jobs — changes there have broader impact.

• On-disk v17 worktree: ecf/scripts/gfs/product/wave/gempak/jgfs_wave_gempak_master.ecf, dev/jobs/JGLOBAL_WAVE_GEMPAK.
• agentcore-mcp-rag MCP (tenant gw_v17): get_change_impact.