IHME Quickstart

This guide covers IHME-specific setup for running Jobmon workflows.

Prerequisites

Before you begin, ensure you have:

  1. Access to the IHME Slurm cluster

  2. A valid project code for job accounting

  3. Conda or a Python environment with Jobmon installed

Environment Setup

Install Jobmon with IHME Configuration

The easiest way to install Jobmon at IHME is with the IHME installer package, which automatically configures the server URL and other IHME-specific settings:

pip install jobmon_installer_ihme

This installs:

  • jobmon_client - the Jobmon client library

  • IHME-specific configuration (server URL, etc.)

Note

If you’re using a shared conda environment that already has Jobmon installed, you may not need to install it yourself. Check with your team.

Running Your First Workflow

  1. SSH to the cluster:

    ssh <username>@<cluster-login-node>

  2. Get an interactive session (required for job submission):

    srun --pty bash

  3. Activate your environment:

    conda activate your_env  # Environment with jobmon_installer_ihme installed

  4. Run your workflow:

    python my_workflow.py

Compute Resources at IHME

Default resources on IHME’s Slurm cluster:

  • Cores: 1

  • Memory: 1GB

  • Runtime: 10 minutes

Specify resources for your tasks:

task = template.create_task(
    name="my_task",
    compute_resources={
        "cores": 2,
        "memory": "10G",
        "runtime": "2h",
        "queue": "all.q",
        "project": "proj_yourproject",
    },
    ...
)

Archive Node Access

To access /snfs1 (J-drive), request archive nodes:

compute_resources={
    "constraints": "archive",
    ...
}

Monitoring at IHME

Jobmon GUI

Access the GUI at: https://jobmon-gui.ihme.washington.edu

  • View workflow progress

  • Investigate failed tasks

  • Check resource usage

CLI Commands

# Check workflow status
jobmon workflow_status -u $USER

# See task details
jobmon workflow_tasks -w <workflow_id>

Getting Help at IHME

For more help, see IHME Support.

Next Steps