Jobs

A job is a way to run non-interactive code in a Databricks cluster. For example, you can run an extract, transform, and load (ETL) workload interactively or on a schedule. You can also run jobs interactively in the notebook UI.

You can create and run a job using the UI, the CLI, or by invoking the Jobs API. You can monitor job run results using the UI, CLI, API, and email alerts. This article focuses on performing job tasks using the UI. For the other methods, see Jobs CLI and Jobs API 2.1.

You can enable orchestration of mulitple tasks in your Databricks workspace. When task orchestration is enabled, a job can consist of one or more tasks, and Databricks manages the task orchestration, cluster management, monitoring, and error reporting for the job. In a workspace without task orchestration enabled, a job consists of a single task. Databricks recommends enabling task orchestration to simplify the creation, management, and monitoring of data and machine learning workflows. See Requirements for information on enabling task orchestration.

You can implement a task in a Databricks notebook, a Delta Live Tables pipeline, or an application written in Scala, Java, or Python. Legacy Spark Submit applications are also supported.

When orchestration of multiple tasks is enabled, you control the execution order of tasks by specifying dependencies between the tasks. You can configure tasks to run in sequence or parallel. The following diagram illustrates a workflow that:

1. Ingests raw clickstream data and performs processing to sessionize the records.

2. Ingests order data and joins it with the sessionized clickstream data to create a prepared data set for analysis.

3. Extracts features from the prepared data.

4. Performs tasks in parallel to persist the features and train a machine learning model.

   

To create your first job orchestrating multiple tasks, see Jobs quickstart.

Important

• You can create jobs only in a Data Science & Engineering workspace or a Machine Learning workspace.
• A workspace is limited to 1000 concurrent job runs. A 429 Too Many Requests response is returned when you request a run that cannot start immediately.
• The number of jobs a workspace can create in an hour is limited to 5000 (includes “run now” and “runs submit”). This limit also affects jobs created by the REST API and notebook workflows.

Requirements

To create jobs that orchestrate multiple tasks, an administrator must enable support in the Databricks admin console.

Create a job

1. Do one of the following:

   • Click Jobs in the sidebar and click .
   • In the sidebar, click Create and select Job from the menu.

   The Tasks tab appears with the create task dialog if orchestration of multiple tasks is enabled. The Task tab appears with the create task dialog if task orchestration is not enabled.

   

2. Replace Add a name for your job… with your job name.

3. If orchestration of multiple tasks is enabled, enter a name for the task in the Task name field.

4. Specify the type of task to run. In the Type drop-down, select Notebook, JAR, Spark Submit, Python, or Pipeline.

   • Notebook: Use the file browser to find the notebook, click the notebook name, and click Confirm.

   • JAR: Specify the Main class. Use the fully qualified name of the class containing the main method, for example, org.apache.spark.examples.SparkPi. Then click Add under Dependent Libraries to add libraries required to run the task. One of these libraries must contain the main class.

     To learn more about JAR tasks, see JAR jobs.

   • Spark Submit: In the Parameters text box, specify the main class, the path to the library JAR, and all arguments, formatted as a JSON array of strings. The following example configures a spark-submit task to run the DFSReadWriteTest from the Apache Spark examples:

     ["--class","org.apache.spark.examples.DFSReadWriteTest","dbfs:/FileStore/libraries/spark_examples_2_12_3_1_1.jar","/dbfs/databricks-datasets/README.md","/FileStore/examples/output/"]
     

     Important

     There are several limitations for spark-submit tasks:

     • You can run spark-submit tasks only on new clusters.
     • Spark-submit does not support cluster autoscaling. To learn more about autoscaling, see Cluster autoscaling.
     • Spark-submit does not support Databricks Utilities. To use Databricks Utilities, use JAR tasks instead.

   • Python: In the Path textbox, enter the URI of a Python script on DBFS or cloud storage; for example, dbfs:/FileStore/myscript.py.

   • Pipeline: In the Pipeline drop-down, select an existing Delta Live Tables pipeline.

   • Python Wheel: In the Package name text box, enter the package to import, for example, myWheel-1.0-py2.py3-none-any.whl. In the Entry Point text box, enter the function to call when starting the wheel. Click Add under Dependent Libraries to add libraries required to run the task.

5. Configure the cluster where the task runs. In the Cluster drop-down, select either New Job Cluster or Existing All-Purpose Clusters.

   • New Job Cluster: Click Edit in the Cluster drop-down and complete the cluster configuration.
   • Existing All-Purpose Cluster: Select an existing cluster in the Cluster drop-down. To open the cluster in a new page, click the icon to the right of the cluster name and description.

   To learn more about selecting and configuring clusters to run tasks, see Cluster configuration tips.

6. You can pass parameters for your task. Each task type has different requirements for formatting and passing the parameters.

   • Notebook: Click Add and specify the key and value of each parameter to pass to the task. You can override or add additional parameters when you manually run a task using the Run a job with different parameters option. Parameters set the value of the notebook widget specified by the key of the parameter. Use task parameter variables to pass a limited set of dynamic values as part of a parameter value.
   • JAR: Use a JSON-formatted array of strings to specify parameters. These strings are passed as arguments to the main method of the main class. See Configure JAR job parameters.
   • Spark Submit task: Parameters are specified as a JSON-formatted array of strings. Conforming to the Apache Spark spark-submit convention, parameters after the JAR path are passed to the main method of the main class.
   • Python: Use a JSON-formatted array of strings to specify parameters. These strings are passed as arguments which can be parsed using the argparse module in Python.
   • Python Wheel: In the Parameters drop-down, select *args (string array) to enter parameters as a JSON-formatted array of strings or select **kwargs (key-value pairs) to enter the key and value of each parameter.

7. To access additional options, including Dependent Libraries, Retry Policy, and Timeouts, click Advanced Options. See Edit a task.

8. Click Create.

9. To optionally set the job’s schedule, click Edit schedule in the Job details panel. See Schedule a job.

10. To optionally allow multiple concurrent runs of the same job, click Edit concurrent runs in the Job details panel. See Maximum concurrent runs.

11. To optionally specify email addresses to receive alerts on job events, click Edit alerts in the Job details panel. See Alerts.

12. To optionally control permission levels on the job, click Edit permissions in the Job details panel. See Control access to jobs.

To add another task when task orchestration is enabled, click below the task you just created.

Run a job

1. Click Jobs in the sidebar.
2. Select a job and click the Runs tab. You can run a job immediately or schedule the job to run later.

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. If you need to make changes to the notebook, clicking Run Now again after editing the notebook will automatically run the new version of the notebook.

Run a job with different parameters

You can use Run Now with Different Parameters to re-run a job with different parameters or different values for existing parameters.

1. Click next to Run Now and select Run Now with Different Parameters or, in the Active Runs table, click Run Now with Different Parameters. Enter the new parameters depending on the type of task.
   • Notebook: You can enter parameters as key-value pairs or a JSON object. You can use this dialog to set the values of widgets.
   • JAR and spark-submit: You can enter a list of parameters or a JSON document. The provided parameters are merged with the default parameters for the triggered run. If you delete keys, the default parameters are used. You can also add task parameter variables for the run.
2. Click Run.

Schedule a job

To define a schedule for the job:

1. Click Edit schedule in the Job details panel and set the Schedule Type to Scheduled.

2. Specify the period, starting time, and time zone. Optionally select the Show Cron Syntax checkbox to display and edit the schedule in Quartz Cron Syntax.

   Note

   • Databricks enforces a minimum interval of 10 seconds between subsequent runs triggered by the schedule of a job regardless of the seconds configuration in the cron expression.
   • You can choose a time zone that observes daylight saving time or UTC. If you select a zone that observes daylight saving time, an hourly job will be skipped or may appear to not fire for an hour or two when daylight saving time begins or ends. To run at every hour (absolute time), choose UTC.
   • The job scheduler is not intended for low latency jobs. Due to network or cloud issues, job runs may occasionally be delayed up to several minutes. In these situations, scheduled jobs will run immediately upon service availability.

3. Click Save.

Pause and resume a job schedule

To pause a job, you can either:

• Click Pause in the Job details panel.
• Click Edit schedule in the Job details panel and set the Schedule Type to Manual (Paused)

To resume a paused job schedule, set the Schedule Type to Scheduled.

View jobs

Click Jobs in the sidebar. The Jobs list appears. The Jobs page lists all defined jobs, the cluster definition, the schedule, if any, and the result of the last run.

You can filter jobs in the Jobs list:

• Using keywords.
• Selecting only the jobs you own.
• Selecting all jobs you have permissions to access. Access to this filter requires that Jobs access control is enabled.

You can also click any column header to sort the list of jobs (either descending or ascending) by that column. The default sorting is by job name in ascending order.

View job runs

1. Click Jobs in the sidebar.
2. In the Name column, click a job name. The Runs tab shows active runs and completed runs.

Databricks maintains a history of your job runs for up to 60 days. If you need to preserve job runs, Databricks recommends that you export results before they expire. For more information, see Export job run results.

View job run details

The job run details page contains job output and links to logs. When orchestration of multiple tasks is enabled, The job run details page contains information about the success or failure of each task in the job run. You can access job run details from the Runs tab for the job. To view job run details from the Runs tab, click the link for the run in the Run column of the Completed Runs (past 60 days) table. To return to the Runs tab for the job, click on the Job ID value.

For a job that orchestrates multiple tasks, click on a task to view task run details, including:

• the cluster that ran the task
  • the Spark UI for the task
  • logs for the task
  • metrics for the task

Click the Job ID value to return to the Runs tab for the job. Click the Job run ID value to return to the job run details.

Export job run results

You can export notebook run results and job run logs for all job types.

Export notebook run results

You can persist job runs by exporting their results. For notebook job runs, you can export a rendered notebook that can later be imported into your Databricks workspace.

To export notebook run results for a job with a single task:

1. On the job detail page, click the View Details link for the run in the Run column of the Completed Runs (past 60 days) table.
2. Click Export to HTML.

To export notebook run results for a job with multiple tasks:

1. On the job detail page, click the View Details link for the run in the Run column of the Completed Runs (past 60 days) table.
2. Click the notebook task to export.
3. Click Export to HTML.

Export job run logs

You can also export the logs for your job run. You can set up your job to automatically deliver logs to DBFS or S3 through the Job API. See the new_cluster.cluster_log_conf object in the request body passed to the Create a new job operation (POST /jobs/create) in the Jobs API.

Edit a job

Some configuration options are available on the job, and other options are available on individual tasks. For example, the maximum concurrent runs can be set on the job only, while parameters must be defined for each task.

To change the configuration for a job:

1. Click Jobs in the sidebar.
2. In the Name column, click the job name.

The side panel displays the Job details. You can change the schedule, cluster configuration, alerts, and maximum number of concurrent runs. You can also edit job permissions if job access control is enabled.

Maximum concurrent runs

The maximum number of parallel runs for this job. Databricks skips the run if the job has already reached its maximum number of active runs when attempting to start a new run. Set this value higher than the default of 1 to perform multiple runs of the same job concurrently. This is useful, for example, if you trigger your job on a frequent schedule and want to allow consecutive runs to overlap with each other, or you want to trigger multiple runs that differ by their input parameters.

Alerts

You can add one or more email addresses to notify when runs of this job begin, complete, or fail:

1. Click Edit alerts.
2. Click Add.
3. Enter the email address to notify and click the check box for each alert type to send to that address.
4. To enter another email address for notification, click Add.
5. If you do not want to receive alerts for skipped job runs, click the check box.
6. Click Confirm.

Integrate these email alerts with your favorite notification tools, including:

• PagerDuty
• Slack
• Amazon SES and SNS

Control access to jobs

Job access control enables job owners and administrators to grant fine-grained permissions on their jobs. Job owners can choose which other users or groups can view the results of the job. Owners can also choose who can manage their job runs (Run now and Cancel run permissions).

See Jobs access control for details.

Edit a task

To set task configuration options:

1. Click Jobs in the sidebar.
2. In the Name column, click the job name.
3. When orchestration of multiple tasks is enabled, click the Tasks tab. If task orchestration is not enabled, click the Task tab.

Task dependencies

When orchestration of multiple tasks is enabled, you can define the order of execution of tasks in a job using the Depends on drop-down. You can set this field to one or more tasks in the job.

Note

Depends on is not visible if the job consists of only a single task.

Configuring task dependencies creates a Directed Acyclic Graph (DAG) of task execution, a common way of representing execution order in job schedulers. For example, consider the following job consisting of four tasks:

• Task 1 is the root task and does not depend on any other task.
• Task 2 and Task 3 depend on Task 1 completing first.
• Finally, Task 4 depends on Task 2 and Task 3 completing successfully.

Databricks runs upstream tasks before running downstream tasks, running as many of them in parallel as possible. The following diagram illustrates the order of processing for these tasks:

Individual task configuration options

Individual tasks have the following configuration options:

Cluster

To configure the cluster where a task runs, click the Cluster drop-down. To learn more about selecting and configuring clusters to run tasks, see Cluster configuration tips.

Note

Each task either runs on an all-purpose cluster or in a new job cluster. You cannot use the same job cluster for more than one task in a job.

Dependent libraries

Dependent libraries will be installed on the cluster before the task runs. You must set all task dependencies to ensure they are installed before the run starts.

To add a dependent library, click Advanced options and select Add Dependent Libraries to open the Add Dependent Library chooser. Follow the recommendations in Library dependencies for specifying dependencies.

Important

If you have configured a library to install on all clusters automatically, or you select an existing terminated cluster that has libraries installed, the job execution does not wait for library installation to complete. If a job requires a specific library, you should attach the library to the job in the Dependent Libraries field.

Task parameter variables

You can pass templated variables into a job task as part of the task’s parameters. These variables are replaced with the appropriate values when the job task runs. You can use task parameter values to pass the context about a job run, such as the run ID or the job’s start time.

When a job runs, the task parameter variable surrounded by double curly braces is replaced and appended to an optional string value included as part of the value. For example, to pass a parameter named MyJobId with a value of my-job-6 for any run of job ID 6, add the following task parameter:

{
  "MyJobID": "my-job-{{job_id}}"
}

The contents of the double curly braces are not evaluated as expressions, so you cannot do operations or functions within double-curly braces. Whitespace is not stripped inside the curly braces, so {{  job_id  }} will not be evaluated.

The following task parameter variables are supported:

Variable	Description	Example value
{{job_id}}	The unique identifier assigned to a job	1276862
{{run_id}}	The unique identifier assigned to a job run	3447843
{{start_date}}	The date a job run started. The format is yyyy-MM-dd in UTC timezone.	2021-02-15
{{start_time}}	The timestamp of the run’s start of execution after the cluster is created and ready. The format is milliseconds since UNIX epoch in UTC timezone, as returned by System.currentTimeMillis().	1551622063030
{{task_retry_count}}	The number of retries that have been attempted to run a task if the first attempt fails. The value is 0 for the first attempt and increments with each retry.

The following task parameter variables are supported when orchestration of multiple tasks is enabled:

Variable	Description	Example value
{{parent_run_id}}	The unique identifier assigned to the run of a job with multiple tasks.	3447835
{{task_key}}	The unique name assigned to a task that’s part of a job with multiple tasks.	“clean_raw_data”

You can set these variables with any task when you Create a job, Edit a job, or Run a job with different parameters.

Timeout

The maximum completion time for a job. If the job does not complete in this time, Databricks sets its status to “Timed Out”.

Retries

A policy that determines when and how many times failed runs are retried. To set the retries for the task, click Advanced options and select Edit Retry Policy.

Note

If you configure both Timeout and Retries, the timeout applies to each retry.

Clone a job

You can quickly create a new job by cloning an existing job. Cloning a job creates an identical copy of the job, except for the job ID. On the job’s page, click More … next to the job’s name and select Clone from the drop-down menu.

Clone a task

You can quickly create a new task by cloning an existing task:

1. On the job’s page, click the Tasks tab.
2. Select the task to clone.
3. Click and select Clone task.

Delete a job

To delete a job, on the job’s page, click More … next to the job’s name and select Delete from the drop-down menu.

Delete a task

To delete a task when orchestration of multiple tasks is enabled:

1. Click the Tasks tab.
2. Select the task to be deleted.
3. Click and select Remove task.

Copy a task path

To copy the path to a task, for example, a notebook path:

1. Click the Tasks tab.
2. Select the task containing the path to copy.
3. Click next to the task path to copy the path to the clipboard.

Best practices

Cluster configuration tips

Cluster configuration is important when you operationalize a job. The following provides general guidance on choosing and configuring job clusters, followed by recommendations for specific job types.

Choose the correct cluster type for your job

• New Job Clusters are dedicated clusters created and started when you run a task and terminate immediately after the task completes. In production, Databricks recommends using new clusters so that each task runs in a fully isolated environment.
• When you run a task on a new cluster, the task is treated as a data engineering (task) workload, subject to the task workload pricing. When you run a task on an existing all-purpose cluster, the task is treated as a data analytics (all-purpose) workload, subject to all-purpose workload pricing.
• If you select a terminated existing cluster and the job owner has Can Restart permission, Databricks starts the cluster when the job is scheduled to run.
• Existing all-purpose clusters work best for tasks such as updating dashboards at regular intervals.

Use a pool to reduce cluster start times

To decrease new job cluster start time, create a pool and configure the job’s cluster to use the pool.

Automatic availability zones

To take advantage of automatic availability zones (Auto-AZ), you must enable it with the Clusters API, setting awsattributes.zone_id = "auto". See Availability zones.

Notebook job tips

Total notebook cell output (the combined output of all notebook cells) is subject to a 20MB size limit. Additionally, individual cell output is subject to an 8MB size limit. If total cell output exceeds 20MB in size, or if the output of an individual cell is larger than 8MB, the run is canceled and marked as failed.

If you need help finding cells near or beyond the limit, run the notebook against an all-purpose cluster and use this notebook autosave technique.

Streaming tasks

Spark Streaming jobs should never have maximum concurrent runs set to greater than 1. Streaming jobs should be set to run using the cron expression "* * * * * ?" (every minute).

Since a streaming task runs continuously, it should always be the final task in a job.

JAR jobs

When running a JAR job, keep in mind the following:

Output size limits

Note

Available in Databricks Runtime 6.3 and above.

Job output, such as log output emitted to stdout, is subject to a 20MB size limit. If the total output has a larger size, the run is canceled and marked as failed.

To avoid encountering this limit, you can prevent stdout from being returned from the driver to Databricks by setting the spark.databricks.driver.disableScalaOutput Spark configuration to true. By default, the flag value is false. The flag controls cell output for Scala JAR jobs and Scala notebooks. If the flag is enabled, Spark does not return job execution results to the client. The flag does not affect the data that is written in the cluster’s log files. Setting this flag is recommended only for job clusters for JAR jobs because it will disable notebook results.

Use the shared SparkContext

Because Databricks is a managed service, some code changes may be necessary to ensure that your Apache Spark jobs run correctly. JAR job programs must use the shared SparkContext API to get the SparkContext. Because Databricks initializes the SparkContext, programs that invoke new SparkContext() will fail. To get the SparkContext, use only the shared SparkContext created by Databricks:

val goodSparkContext = SparkContext.getOrCreate()
val goodSparkSession = SparkSession.builder().getOrCreate()

There are also several methods you should avoid when using the shared SparkContext.

• Do not call SparkContext.stop().
• Do not call System.exit(0) or sc.stop() at the end of your Main program. This can cause undefined behavior.

Use try-finally blocks for job clean up

Consider a JAR that consists of two parts:

• jobBody() which contains the main part of the job.
• jobCleanup() which has to be executed after jobBody() whether that function succeeded or returned an exception.

As an example, jobBody() may create tables, and you can use jobCleanup() to drop these tables.

The safe way to ensure that the clean up method is called is to put a try-finally block in the code:

try {
  jobBody()
} finally {
  jobCleanup()
}

You should not try to clean up using sys.addShutdownHook(jobCleanup) or the following code:

val cleanupThread = new Thread { override def run = jobCleanup() }
Runtime.getRuntime.addShutdownHook(cleanupThread)

Due to the way the lifetime of Spark containers is managed in Databricks, the shutdown hooks are not run reliably.

Configure JAR job parameters

You pass parameters to JAR jobs with a JSON string array. See the spark_jar_task object in the request body passed to the Create a new job operation (POST /jobs/create) in the Jobs API. To access these parameters, inspect the String array passed into your main function.

Library dependencies

The Spark driver has certain library dependencies that cannot be overridden. These libraries take priority over any of your libraries that conflict with them.

To get the full list of the driver library dependencies, run the following command inside a notebook attached to a cluster of the same Spark version (or the cluster with the driver you want to examine).

%sh
ls /databricks/jars

Manage library dependencies

A good rule of thumb when dealing with library dependencies while creating JARs for jobs is to list Spark and Hadoop as provided dependencies. On Maven, add Spark and Hadoop as provided dependencies, as shown in the following example:

<dependency>
  <groupId>org.apache.spark</groupId>
  <artifactId>spark-core_2.11</artifactId>
  <version>2.3.0</version>
  <scope>provided</scope>
</dependency>
<dependency>
  <groupId>org.apache.hadoop</groupId>
  <artifactId>hadoop-core</artifactId>
  <version>1.2.1</version>
  <scope>provided</scope>
</dependency>

In sbt, add Spark and Hadoop as provided dependencies, as shown in the following example:

libraryDependencies += "org.apache.spark" %% "spark-core" % "2.3.0" % "provided"
libraryDependencies += "org.apache.hadoop" %% "hadoop-core" % "1.2.1" % "provided"

Tip

Specify the correct Scala version for your dependencies based on the version you are running.