Autosubmit mHM workflow with mHM test domains data

This repository contains the configuration for an Autosubmit experiment that runs mHM, the mesoscale hydrological model, using the mHM test domain data in a workflow. A Docker container is used to install the dependencies and run mHM.

The two images below were created using the workflow in this repository, and running the plot.py script (it uses Xarray and Matplotlib to parse the NetCDF test output).

Prerequisites

In order to run the workflow you will need the following prerequisites:

1. Internet connection
2. Linux
3. Autosubmit 4.0.98 or greater (e.g. pip install autosubmit==4.0.98)
4. Docker & Singularity 3.11.x or greater for running the container
5. ImageMagick's convert to generate the GIF animation using the PNG files produced by plot.py

For a list of software used, besides mHM, see the Dockerfile.

Build the containers

To build the Docker container image first, use this command.

sudo docker build --no-cache=true --tag "auto-mhm-test-domains/mhm:v5.12.1.dev228" .

Now, to build the Singularity container, using the Docker container image, use this command.

sudo singularity build --force mhm.sif docker-daemon://auto-mhm-test-domains/mhm:v5.12.1.dev228

Now copy the created container mhm.sif into the location specified in conf/mhm.yml.

NOTE: The workflow expects the mhm.sif Singularity file to exist on each platform.

Workflow

An Autosubmit experiment created with autosubmit expid -H <YOUR_PLATFORM> -d "mHM" -min -repo https://github.com/kinow/auto-mhm-test-domains.git -git_branch master -conf conf/bootstrap when run will clone this Git repository, and prepare the Docker & Singularity containers, transfer all the required data over to the chosen platform (it can be local for testing) and execute the mHM model for each start date (as the mHM simulation period date). The last tasks in the workflow will plot the data before cleaning the simulation logs and files.

The plots are copied back to the local workflow folder, and everything can be used to package an RO-Crate.

Running

You will need an Autosubmit experiment first, so that you can import the configuration files from this Git repository. Run the following command to create a new Autosubmit experiment.

autosubmit expid \
    --description "Autosubmit mHM test domains" \
    --minimal_configuration \
    --git_as_conf conf/bootstrap \
    --git_repo https://github.com/kinow/auto-mhm-test-domains.git \
    --git_branch master

This will create a new experiment, using SSH connections to the “local” platform. On the first execution of autosubmit create, it will clone the Git repository specified and load the configuration from the subdirectory specified in the git_as_conf parameter.

NOTE: The output of autosubmit expid contains the ID of an experiment. Replace <EXPID> by that value in the next commands.

The platforms.yml file in this repository contains placeholders like <USER>, <REMOTE_HOST>, etc., that must be filled in before the workflow can be used. Alternatively, you can create the following file with your platform configuration (replace the values by your user account information).

# File: ~/.config/autosubmit/platforms.yml

PLATFORMS:
  # You can use a different name, just adjust your `autosubmit expid`
  # command and `conf/platforms.yml`, and this other platforms.yml
  # files.
  REMOTE:
    TYPE: ps
    HOST: localhost
    ADD_PROJECT_TO_HOST: false
    SCRATCH_DIR: /tmp/scratch/
    PROJECT: mhm-project
    USER: kinow

NOTE: The SCRATCH_DIR and PROJECT directories must already exist or be created before following the next steps.

The next command is to prepare the experiment workflow (i.e. parse and validate its configuration and produce a workflow graph, prepare jobs and scripts, etc.):

autosubmit create <EXPID>

If everything goes well you should see the workflow graph plot appear on your screen (if you have an X server running). Close the PDF and now run the workflow with the following command:

autosubmit run <EXPID>

That will execute the complete workflow of your Autosubmit experiment. If you used nohup, or if you have another command-line terminal, you can monitor the execution of the workflow with the following commands:

# plot a new PDF with the progress of your workflow
autosubmit monitor <EXPID>
# print the output logs of your workflow
autosubmit cat-log --file o <EXPID>
# print the error logs of your workflow, in `tail -f` mode
autosubmit cat-log --file o --mode e <EXPID>

NOTE: Autosubmit commands produce log files on disk for traceability. You can increase the log levels with the -lf (log file) and lc (log console) parameter flags.

And if you are using a version of Autosubmit that supports RO-Crate, you can create an archive with the provenance metadata file, the workflow configuration, plots, logs, and other traces included.

autosubmit archive --rocrate <EXPID>

NOTE: If you use RO-Crate, you will have to provide a JSON-LD file with extra metadata about your experiment and workflow to have a more complete RO-Crate file (using the Workflow Run Crate Profile).

RO-Crate

This is an example RO-Crate configuration file for this Autosubmit workflow. You can add this setting anywhere, like <EXPID>/conf/rocrate.yml.

ROCRATE:
  INPUTS:
    # Add the extra configuration keys to be exported.
    # This one is from the mhm.yml file, we export everything
    # as inputs.
    - "MHM"
  OUTPUTS:
    - "*/*.gif"
  PATCH: |
    {
      "@graph": [
        {
          "@id": "./",
          "license": "Apache-2.0",
          "creator": {
            "@id": "https://orcid.org/0000-0001-8250-4074"
          },
          "publisher": {
            "@id": "https://ror.org/05sd8tv96"
          }
        },
        {
          "@id": "#create-action",
          "@type": "CreateAction",
          "name": "Run mHM",
          "instrument": { "@id": "workflow.yml" },
          "agent": { "@id": "https://orcid.org/0000-0001-8250-4074" }
        },
        {
          "@id": "ro-crate-metadata.json",
          "author": [
            {
              "@id": "https://orcid.org/0000-0001-8250-4074"
            }
          ]
        },
        {
          "@id": "https://orcid.org/0000-0001-8250-4074",
          "@type": "Person",
          "affiliation": {
              "@id": "https://ror.org/05sd8tv96"
          },
          "contactPoint": {
              "@id": "mailto: [email protected]"
          },
          "name": "Bruno P. Kinoshita"
        },
        {
            "@id": "mailto: [email protected]",
            "@type": "ContactPoint",
            "contactType": "Author",
            "email": "[email protected]",
            "identifier": "[email protected]",
            "url": "https://orcid.org/0000-0001-8250-4074"
        },
        {
            "@id": "https://ror.org/05sd8tv96",
            "@type": "Organization",
            "name": "Barcelona Supercomputing Center"
        }
      ]
    }

Troubleshooting

If you have issues running the workflow that are related to the containers, you can try re-building without the cache by using the docker build --no-cache flag.

Another option to troubleshoot issues with the containers is to use the final container image, mhm.sif, directly in the command-line, without the workflow.

For example:

singularity run mhm.sif mhm-download -b develop -d 1 -p /tmp/data

License

This workflow is licensed under the GNU General Public License V3. You can find the complete license at the file LICENSE.txt.

The license applies only to the workflow configuration and code, and not to the rest of the code and tools used.

Please, note that:

• Autosubmit is licensed under the GNU General Public License v3.
• mHM is licensed under the GNU General Public License v3.