Trigger a single job run

Use the Run now button to trigger a single job run.

You can also choose to Run now with different parameters to override default job parameters.

Note

Databricks recommends using the default settings that enforce a single concurrent run for each job. If your workload requires multiple concurrent runs of a job, see Configure maximum concurrent runs.

Run a job immediately

To run the job immediately, click .

Tip

You can perform a test run of a job with a notebook task by clicking Run now. Click Run now again after editing the notebook to run the latest version of the notebook.

Run a job with different settings

Run now with different settings allows you to choose which tasks to run, different parameters to run with, to override default values for existing parameters, and to run Performance optimized serverless workloads.

1. Click the blue caret next to Run now and select Run now with different settings or, in the Active Runs table, click Run now with different settings.
2. (Optional) Select or clear the tasks to run by clicking them in the graph. To run a task together with its upstream or downstream tasks, see Run a subset of tasks and their dependencies.
3. (Optional) Enter the new job parameters as key-value pairs. See Configure job parameters.
4. (Optional) Change the Performance optimized setting. See Run your Lakeflow Jobs with serverless compute for workflows.
5. Click Run.

Deselecting a task only affects this run. To skip a task on every run without deleting it, disable the task in the job settings. See Disabled tasks in Lakeflow Jobs.

Note

The Run now with different parameters dialog contains an option to Switch to legacy parameters if your current job only has task parameters and not job parameters defined. This legacy behavior is no longer recommended.

Run a subset of tasks and their dependencies

When you run with different parameters, you can select a subset of tasks to run. You can optionally expand your selection to include each task's dependencies: its upstream tasks (the tasks that come before, that it depends on) or its downstream tasks (the tasks that come after it, and depend on it).

When you select tasks by key in the API or CLI, the + modifier controls this expansion:

Task key	Tasks that run
my_task	Only my_task.
+my_task	my_task and all of its upstream tasks.
my_task+	my_task and all of its downstream tasks.
+my_task+	my_task and both its upstream and downstream tasks.

The + expansion excludes disabled tasks. To include disabled tasks, you must select them explicitly. See Disabled tasks in Lakeflow Jobs.

Start a partial run from the UI, the REST API, the Azure Databricks CLI, the SDKs, or Declarative Automation Bundles:

UI

1. Open the Run now with different settings dialog and click Deselect all. See Run a job with different settings.
2. Click the task you want to run in the graph.
3. Hover over the selected task. In the menu bar above the task, click to also run its upstream tasks or to also run its downstream tasks. Click both to include upstream and downstream tasks. These buttons are equivalent to the +my_task, and my_task+ syntax.
4. Click Run.

REST API

Pass the task keys in the only field of a jobs/run-now request. If you omit only, the job runs all of its tasks.

The following request runs task_1 with its upstream tasks and task_2 with its downstream tasks:

POST /api/2.2/jobs/run-now

{
  "job_id": 11223344,
  "only": ["+task_1", "task_2+"]
}

The SDKs accept the same only field.

CLI

Pass the task keys to the --only flag of databricks bundle run as a comma-separated list. The same flag applies when you run a job defined with Declarative Automation Bundles:

databricks bundle run my_job --only +task_1,task_2+

Manual triggers and continuous jobs

Continuous jobs should always keep the default setting of a single concurrent run. As such, if a continuous trigger is active, the Run now button is replaced by a Restart run button.

If you pause a continuous trigger, the Run now button becomes available.

If any run is active when a continuous trigger is resumed, the job scheduler waits until that run completes to trigger a new run.