Skip to content

Run SOMISANA forecast models #613

Run SOMISANA forecast models

Run SOMISANA forecast models #613

Workflow file for this run

name: Run SOMISANA forecast models
on:
workflow_dispatch:
inputs:
run_date:
description: Run date (default = today). yyyymmdd
required: false
default:
# commenting the schedule for now until we are ready to deploy this workflow operationally
schedule:
- cron: '0 02 * * *' # 0400 SAST
# Many of the env variables reference the current branch,
# which is set dynamically in the branch-ref job below
# So the env variables are mostly set in the envs job below the branch-ref job
env:
REGISTRY: ghcr.io
jobs:
# Dynamically set the branch ref to the currently executing branch
branch-ref:
runs-on: ubuntu-latest
outputs:
value: ${{ steps.BRANCH_REF.outputs.value }}
steps:
- name: Set the BRANCH_REF
id: BRANCH_REF
run: |
echo "value=${GITHUB_REF##*/}" >> $GITHUB_OUTPUT
# Set the environment variables using the current branch reference
envs:
needs: [branch-ref]
runs-on: ubuntu-latest
env:
RUN_DATE: ${{ inputs.run_date }}
outputs:
BRANCH_REF: ${{ needs.branch-ref.outputs.value }}
TOOLKIT_IMAGE: ${{ steps.ENVS.outputs.TOOLKIT_IMAGE }}
MODEL_RUN_DATE: ${{ steps.ENVS.outputs.MODEL_RUN_DATE }}
EEZ_DATA_DIR: ${{ steps.ENVS.outputs.EEZ_DATA_DIR }}
steps:
- name: Configure run date
id: run_date
run: |
echo "yyyymmdd=${RUN_DATE:=$(date +'%Y%m%d')}" >> $GITHUB_OUTPUT
- name: Set envs
id: ENVS
run: |
echo "TOOLKIT_IMAGE=${{ github.repository }}_toolkit_${{ needs.branch-ref.outputs.value }}" >> $GITHUB_OUTPUT
echo "MODEL_RUN_DATE=${{ steps.run_date.outputs.yyyymmdd }}" >> $GITHUB_OUTPUT
echo "EEZ_DATA_DIR=/home/runner/somisana/global_data/${{ steps.run_date.outputs.yyyymmdd }}/${{ needs.branch-ref.outputs.value }}" >> $GITHUB_OUTPUT # Think it would have been better to have $BRANCH_REF after somisana i.e. somisana/$BRANCH_REF but doing it this way to be consistent with how the model run directories were set up. Maybe we can change it later...
# The SOMISANA toolkit is a suite of scripts to facilitate
# running SOMISANA models - for example the Algoa Bay Forecast.
# In the context of a testing environment, where updates to these
# scripts is part of the development workflow, the toolkit needs
# to be compiled on every run
toolkit:
needs: [envs]
runs-on: ubuntu-latest
env:
BRANCH_REF: ${{ needs.envs.outputs.BRANCH_REF }}
TOOLKIT_IMAGE: ${{ needs.envs.outputs.TOOLKIT_IMAGE }}
SHA: sha-${{ github.sha }}
outputs:
image: ${{ env.REGISTRY }}/${{ steps.lowercase.outputs.image_name }}
steps:
- name: Get image name lowercase
id: 'lowercase'
run: |
IMAGE_NAME_LOWER=$(echo $TOOLKIT_IMAGE | tr '[:upper:]' '[:lower:]')
echo "image_name=$IMAGE_NAME_LOWER" >> $GITHUB_OUTPUT
- name: Check out source code
uses: actions/checkout@main # change to main
with:
ref: ${{ env.BRANCH_REF }}
- name: Log in to the Container registry
uses: docker/login-action@master
with:
registry: ${{ env.REGISTRY }}
username: ${{ github.actor }}
password: ${{ secrets.GITHUB_TOKEN }}
- name: Check if image with sha tag already exists
id: check-image
run: |
set -e
SHA_IMAGE=$(docker images -q ${REGISTRY}/${{ steps.lowercase.outputs.image_name }}:${SHA} || true)
if [ -z "$SHA_IMAGE" ]; then
echo "image_exists=false" >> $GITHUB_OUTPUT
else
echo "image_exists=true" >> $GITHUB_OUTPUT
fi
- name: Extract metadata (tags, labels) for Docker
id: meta
uses: docker/metadata-action@master
with:
images: ${{ env.REGISTRY }}/${{ env.TOOLKIT_IMAGE }}
tags: |
latest
${{ env.SHA }}
- name: Build and push
uses: docker/build-push-action@master
with:
context: toolkit
push: true
tags: ${{ steps.meta.outputs.tags }}
labels: ${{ steps.meta.outputs.labels }}
# Download environmental data used for model boundaries and surface forcing.
# This job is retried if it fails, since the failure can be due to upstream errors
# These downloads are used to create forcing files that are fed as input to the CROCO model
# => marine.copernicus.eu: This is ocean data that forms the lateral boundaries of our model run
# => ncei.noaa.gov: This is atmospheric data used to create surface boundary conditions for our model run
download-global-data:
needs: [toolkit, envs]
runs-on: somisana
env:
EEZ_DATA_DIR: ${{ needs.envs.outputs.EEZ_DATA_DIR }}
MODEL_RUN_DATE: ${{ needs.envs.outputs.MODEL_RUN_DATE }}
SHA: sha-${{ github.sha }}
BRANCH_REF: ${{ needs.envs.outputs.BRANCH_REF }}
steps:
- name: Clean /home/runner/somisana/global_data/
run: >-
find \
/home/runner/somisana/global_data/* \
-maxdepth 0 \
-type d \
-ctime +5 \
-exec \
rm \
-rf {} \;
- name: Configure no. retries
id: retries
run: |
if [ "$BRANCH_REF" = "stable" ]
then
echo "RETRIES=30" >> $GITHUB_OUTPUT
else
echo "RETRIES=1" >> $GITHUB_OUTPUT
fi
- name: Source ~/.bashrc (for NVM - required for non-interactive shells)
continue-on-error: true # Maybe Node.js exists anyway
run: |
source ~/.bashrc
nvm use 19.6.0
- name: Create ${{ env.EEZ_DATA_DIR }} directory
continue-on-error: true
run: |
# rm -rf ${{ env.EEZ_DATA_DIR }} # not sure why we need this... just commenting for now due to issues with nomads, and our code doesn't pick up when randomly don't get downloaded with status code 404. So just re-running the job manually until I have time to fix it properly. And we need the previously downloaded files, otherwise it fails again
mkdir -p ${{ env.EEZ_DATA_DIR }}
chown -R :runners ${{ env.EEZ_DATA_DIR }}
chmod -R 774 ${{ env.EEZ_DATA_DIR }}
- name: Download GFS
uses: nick-fields/retry@master
with:
timeout_minutes: 30 # Script is considered failed if this limit is reached
retry_wait_seconds: 300 # Wait 5 minutes and try again
max_attempts: ${{ steps.retries.outputs.RETRIES }}
retry_on: any
warning_on_retry: true
shell: bash
continue_on_error: false
on_retry_command: rm -f ${{ env.EEZ_DATA_DIR }}/*grb*
command: >-
docker run \
--rm \
-v ${{ env.EEZ_DATA_DIR }}/:/tmp/somisana/current \
${{ needs.toolkit.outputs.image }}:${{ env.SHA }} \
download \
--provider gfs \
--workdir /tmp/somisana/current \
--matlab-env /tmp/somisana/current/.env \
--download-date ${{ env.MODEL_RUN_DATE }} \
--domain 11,36,-39,-25
- name: Download Mercator
uses: nick-fields/retry@master
with:
timeout_minutes: 60 # Script is considered failed if this limit is reached
retry_wait_seconds: 300 # Wait 5 minutes and try again
max_attempts: ${{ steps.retries.outputs.RETRIES }}
retry_on: any
warning_on_retry: true
shell: bash
continue_on_error: false
on_retry_command: rm -f ${{ env.EEZ_DATA_DIR }}/*mercator*
command: >-
docker run \
--rm \
-v ${{ env.EEZ_DATA_DIR }}/:/tmp/somisana/current/ \
-e COPERNICUS_USERNAME=${{ secrets.COPERNICUS_USERNAME }} \
-e COPERNICUS_PASSWORD=${{ secrets.COPERNICUS_PASSWORD }} \
${{ needs.toolkit.outputs.image }}:${{ env.SHA }} \
download \
--provider mercator \
--workdir /tmp/somisana/current \
--download-date ${{ env.MODEL_RUN_DATE }} \
--domain 11,36,-39,-25
# now do the model specific steps
# currently this is done by calling different model reusable workflows in series
# In future we should make a generic model reusable workflow which takes the model name as an input
# Then the model runs can be carried out using a matrix strategy
# (there's an example of this here: https://docs.github.com/en/actions/using-workflows/reusing-workflows#example-reusable-workflow)
call_run_algoa-bay-forecast:
needs: [envs,toolkit,download-global-data]
uses: ./.github/workflows/run_algoa-bay-forecast.yml
with:
BRANCH_REF: ${{ needs.envs.outputs.BRANCH_REF }}
MODEL_RUN_DATE: ${{ needs.envs.outputs.MODEL_RUN_DATE }}
TOOLKIT_IMAGE: ${{ needs.toolkit.outputs.image }}:sha-${{ github.sha }}
EEZ_DATA_DIR: ${{ needs.envs.outputs.EEZ_DATA_DIR }}
secrets: inherit
call_run_false-bay-forecast:
needs: [envs,toolkit,download-global-data]
uses: ./.github/workflows/run_false-bay-forecast.yml
with:
BRANCH_REF: ${{ needs.envs.outputs.BRANCH_REF }}
MODEL_RUN_DATE: ${{ needs.envs.outputs.MODEL_RUN_DATE }}
TOOLKIT_IMAGE: ${{ needs.toolkit.outputs.image }}:sha-${{ github.sha }}
EEZ_DATA_DIR: ${{ needs.envs.outputs.EEZ_DATA_DIR }}
secrets: inherit