Create and run Jobs

Learn about the many flexible ways that you can launch Domino Jobs depending on your needs. Domino Jobs provide a structured approach to run code files organizing the necessary hardware, compute environment, data resources, and security protocols to create a highly reproducible execution environment.

Run a Job in the UI

You can start a Job interactively from multiple locations in the Domino UI.

Jobs dashboard in a Project

To run a Job from the Job dashboard in your Project:

In your project, navigate to the Jobs section. The Jobs dashboard shows all Jobs for your Project, sorted into tabs by Job status.

Note
You can also use the Quick Action menu to create a Job. To do this, click the + next to Jobs in the navigation pane.
Click Run to start a new Job.
In the Start Run window, configure the Job.
Enter the File Name of the script you want Domino to execute, followed by any arguments for the script.
Choose the Compute Environment for the Job and its specific version. By default, the Job will use the latest version of your environment.

Tip
Domino recommends that you use the revision that was set as active by the Project owner. If you select another revision, you can see a Not Recommended warning.

Note
Restricted Projects may limit your choice of environments.
If your administrator has enabled volume provisioning recommendations, you can select a volume size automatically recommended by Domino, instead of the default volume size.

Note
In a Git-based Project, the first Job you launch always uses the volume size configured in your Project settings. Subsequent Job launches receive volume provisioning recommendations.
Click Next and select the Hardware Tier to use for this Job.

In a Domino Nexus deployment, hardware tiers in the dropdown are grouped by data plane, such as GCP below. The Local data plane corresponds to running the execution in the Domino control plane cluster.
If necessary, attach a compute cluster to your execution. To learn more about clusters, see the following:

Domino shows you what data will be mounted in your execution based on the selected data plane.

Tip	If you want to use a data mount that is not available, note which data planes it is available in, then return to the first step in the modal to select a hardware tier in the corresponding data plane.

You can also configure your Job to create new Dataset Snapshots to mount upon execution completion. Created snapshots are shown under the details tab for a Job.

Data in selected data plane

Note	This feature requires `AllowDatasetSnapshotsOnExecutionCompletion` to be enabled. See the create a snapshot guide for more details.

Project files page

You can start a Job directly from the Project Code page.

Go to your Project > Code > Find the file you want to run.
Click the gear icon > Run.
Go to the file you want to run and click the gear icon at the end of the row in the files list, then click Run.
Use the window to configure the Run. This window has additional options to set up recurring scheduled Jobs and select different hardware tiers.

Title

In this field, enter a name for the Job. This name will be the label for the Job in the Jobs dashboard.

Parameters

If there are any arguments or parameters you’d like to pass to your script, enter them here.

Hardware tier

This dropdown list lets you set the hardware tier used by the Job.

Schedule to repeat

In this section, specify if you want the run to execute only once, or to repeat on a schedule.

Publish after complete

Check this if you want Domino to republish a Domino endpoint once the Job completes successfully.

Launcher UI

A Launcher is a web form that rests on top of a script that Domino can execute in a Job. You can use a Launcher to pass arguments to your script from UI fields in a customized web form.

Run a Job from the CLI

From your workstation, you can start Jobs through the Domino CLI. You must first install the Domino CLI on your system.

After you have logged in and opened a Project, you can start Jobs with:

`domino run <filename>`

Examples:

`domino run calculateStats.r`
`domino run runModel.m model1 .05`
`domino run --wait runModel.m model1 .05`
`domino run --direct "pip freeze | grep pandas"`

Domino Python SDK

You can use the Domino Python SDK to start and monitor Jobs in Domino either from a tool like a Jupyter Notebook or VS Code or from within another Domino Job to chain together multiple Jobs in a more complex workflow.

Write code to connect to Domino and start a Run using the following example:

import os

from domino import Domino

domino = Domino(
    "marks/quick-start",
    api_key = os.environ["DOMINO_USER_API_KEY"],
    host = os.environ["DOMINO_API_HOST"],
)

# Blocking: this will start the run and wait for the run to finish before returning the status of the run
domino_run = domino.runs_start_blocking(
    ["main.py", "a", "b", "9"], title = "Started from Python API example code"
)
print(domino_run)


# Non-blocking: this will start the run and print the Run ID and check the status of the run *once*
# domino_run = domino.runs_start(["main.py", "a", "b", "9"],
#                                title = "Started from Python API example code")
# print(domino_run)
# run_status = domino.runs_status(domino_run.get("runId"))
# print(run_status)

Note	When using the Python SDK from within a Domino Workspace or Job, the environment variables needed for authentication are automatically configured. If you would like to set up your workstation outside Domino, learn more about how to programmatically access Domino

Run a Job from the Domino API

You can use the Domino API to start a Job with a POST request to:

https://<domino-url>/v1/projects/<username>/<project-name>/runs

For more information about how to send valid configuration data, see the API docs.

Python and R Domino API bindings

The Domino API is also available in language-specific wrappers.

Use the dominodatalab Python library and domino R package to call the Domino API in your preferred language.

Schedule a Job

Use Domino to schedule Jobs in advance, and set them to execute on a regular schedule. These can be useful when you have a data source that is updated regularly.

Organize and manage

Jobs Data science is an experimental science - it is common for a Project to require many Jobs before you’ve achieved your goals. Domino allows you to tag and rename Jobs to help you and your team track and manage Project work history effectively.

Tag Jobs

Tagged Jobs facilitates better search and discoverability in the Jobs dashboard.

It’s best practice to start tagging Jobs from the start of a Project. For your tagging strategy, think about groups of Jobs you and collaborators would like to view in the Jobs dashboard.

For example, tagging your LLM pre-training Jobs differently than the instruction fine-tuning Jobs might provide for better search and visualization.

From the navigation pane, click Jobs.
Select the Jobs you want to tag, then click the Tag icon.

Rename Jobs

Rename Jobs to clarify their purpose or highlight the importance of a particular Job.

From the navigation pane, click Jobs.
Click the Edit button next to the Job name.

Stop a Job

From the navigation pane, click Jobs.
Go to the Active tab.
Select the checkboxes for the Jobs to stop and click the Stop icon.

Re-run a Job

To re-run a job:

From the Jobs page, select the Run that you want.
Select either Re-Run with Original Version or Re-run with the Latest version

Re-run with Original Version runs the Job with the same input arguments and Project files as the original run.

Re-run with Latest Version runs the Job with the same input arguments used previously, while using the latest versions of your files.

Note	When you work with a Git-based Project, Domino pulls the latest files from the configured branch. If the Git repository is associated with a specific commit, then Domino pulls from that commit.

Note	In restricted Projects, Job re-runs use the current restricted revision of the environment. This revision may be different from the revision that was used to run the Job previously.

Next steps

Learn how to schedule Jobs and view Job results.