Del via

Databricks Asset Bundle deployment modes

  • Artikkel

This article describes the syntax for Databricks Asset Bundle deployment modes. Bundles enable programmatic management of Azure Databricks workflows. See What are Databricks Asset Bundles?

In CI/CD workflows, developers typically code, test, deploy, and run solutions in various phases, or modes. For example, the simplest set of modes includes a development mode for pre-production validation, followed by a production mode for validated deliverables. Databricks Asset Bundles provides an optional collection of default behaviors that correspond to each of these modes. To use these behaviors for a specific target, set a mode or configure presets for a target in the targets configuration mapping. For information on targets, see bundle configuration targets mapping.

Development mode

To deploy your bundle in development mode, you must first add the mode mapping, set to development, to the intended target. For example, this target named dev is treated as a development target:

targets:
  dev:
    mode: development

Deploying a target in development mode by running the databricks bundle deploy -t <target-name> command implements the following behaviors, which can be customized using presets:

  • Prepends all resources that are not deployed as files or notebooks with the prefix [dev ${workspace.current_user.short_name}] and tags each deployed job and pipeline with a dev Azure Databricks tag.
  • Marks all related deployed Delta Live Tables pipelines as development: true. See Use development mode to run pipeline updates.
  • Enables the use of --compute-id <cluster-id> in related calls to the bundle deploy command, which overrides any and all existing cluster definitions that are already specified in the related bundle configuration file. Instead of using --compute-id <cluster-id> in related calls to the bundle deploy command, you can set the compute_id mapping here, or as a child mapping of the bundle mapping, to the ID of the cluster to use.
  • Pauses all schedules and triggers on deployed resources such as jobs or quality monitors. Unpause schedules and triggers for an individual job by setting schedule.pause_status to UNPAUSED.
  • Enables concurrent runs on all deployed jobs for faster iteration. Disable concurrent runs for an individual job by setting max_concurrent_runs to 1.
  • Disables the deployment lock for faster iteration. This lock prevents deployment conflicts which are unlikely to occur in dev mode. Re-enable the lock by setting bundle.deployment.lock.enabled to true.

Production mode

To deploy your bundle in production mode, you must first add the mode mapping, set to production, to the intended target. For example, this target named prod is treated as a production target:

targets:
  prod:
    mode: production

Deploying a target in production mode by running the databricks bundle deploy -t <target-name> command implements the following behaviors:

  • Validates that all related deployed Delta Live Tables pipelines are marked as development: false.

  • Validates that the current Git branch is equal to the Git branch that is specified in the target. Specifying a Git branch in the target is optional and can be done with an additional git property as follows:

    git:
  branch: main

    This validation can be overridden by specifying --force while deploying.

  • Databricks recommends that you use service principals for production deployments. You can enforce this by setting run_as to a service principal. See Manage service principals and Specify a run identity for a Databricks Asset Bundles workflow. If you do not use service principals, then note the following additional behaviors:

    • Validates that artifact_path, file_path, root_path, or state_path mappings are not overridden to a specific user.
    • Validates that the run_as and permissions mappings are specified to clarify which identities have specific permissions for deployments.

  • Unlike the preceding behavior for setting the mode mapping to development, setting the mode mapping to production does not allow overriding any existing cluster definitions that are specified in the related bundle configuration file, for instance by using the --compute-id <cluster-id> option or the compute_id mapping.

Custom presets

Databricks Asset Bundles supports configurable presets for targets, which allows you to customize the behaviors for targets. The available presets are listed in the following table:

Preset Description
name_prefix The prefix string to prepend to resource names.
pipelines_development Whether or not the pipeline is in development mode. Valid values are true or false.
trigger_pause_status A pause status to apply to all triggers and schedules. Valid values are PAUSED or UNPAUSED.
jobs_max_concurrent_runs The number of maximum allowed concurrent runs for jobs.
tags A set of key:value tags that apply to all resources that support tags, which includes jobs and experiments. Databricks Asset Bundles do not support tags for the schema resource.

Note

If both mode and presets are set, presets override the default mode behavior, and settings of individual resources override the presets. For example, if a schedule is set to UNPAUSED, but the trigger_pause_status preset is set to PAUSED, the schedule is unpaused.

The following example shows a custom presets configuration for the target named dev:

targets:
  dev:
    presets:
      name_prefix: "testing_"      # prefix all resource names with testing_
      pipelines_development: true  # set development to true for pipelines
      trigger_pause_status: PAUSED # set pause_status to PAUSED for all triggers and schedules
      jobs_max_concurrent_runs: 10 # set max_concurrent runs to 10 for all jobs
      tags:
        department: finance