Before you deploy¶
Read this page first. It explains deploy order, who deploys what across environments, and where bundle files land in the workspace — before you use the local or CI/CD guides.
Deploy order¶
Pipeline bundles depend on a deployed framework at a known framework_source_path. Deploy the Framework Bundle first, then Pipeline Bundle(s).
Framework Bundle → Pipeline Bundle(s) → pipelines run in the workspace.
Both bundle types use the same Databricks CLI workflow — authenticate, validate, deploy — but differ in bundle root, workspace path, and variables:
Framework Bundle — framework source under workspace files (for example
.bundle/.../files/src)Pipeline Bundle — your data flows and Spark Declarative Pipeline definitions; requires the framework already deployed at
framework_source_path
When building pipeline bundles, confirm the framework is available first — see Building a Pipeline Bundle.
Who deploys what¶
Bundle |
Typical owner |
Dev |
Test / staging / prod |
|---|---|---|---|
Framework |
Platform team (central) or federated platform per domain |
CI/CD as an owning service principal (SP) — same pattern as higher tiers |
CI/CD as an owning SP; versioned promote to a shared or secured workspace path |
Pipeline |
Data / domain teams |
Developer deploys from laptop during inner loop |
CI/CD as an owning SP after merge / approval |
Central vs federated framework deployment is an organizational choice. The docs describe both without prescribing one model.
Framework Bundle — in most organizations the framework is deployed the same way in every environment (dev through prod): platform CI/CD authenticates as an owning service principal, not as an individual developer. Solo dev or POC work may still deploy the framework to a developer’s workspace files for experimentation; that is the exception, not the standard.
Pipeline Bundle — deployment identity usually depends on the environment:
Dev — developers deploy locally from their machine while authoring and testing data flows
All other environments — CI/CD deploys as an owning SP, consistent with the framework pattern above
Service principals and workspace paths¶
When deploying as an SP (framework in all environments; pipeline bundles outside dev), the SP is the owner of the bundle in that workspace. CI/CD jobs authenticate as that SP (or a deployment SP permitted to act on its behalf) so ownership, ACLs, and audit trails stay consistent across promotions.
Bundle files can land in one of two workspace locations:
SP workspace files — default DAB layout under the SP’s user folder, for example
/Workspace/Users/<sp-name>/.bundle/.../files/srcDesignated secured folder — a shared, ACL-controlled path under workspace files (for example
/Workspace/Shared/platform/lakeflow_framework/...) that platform teams manage and pipeline bundles reference viaframework_source_path
Which path you use is an org choice. The framework is commonly deployed to a secured shared folder or the platform SP’s .bundle path in every environment; pipeline bundles in dev typically use the developer’s workspace files, then switch to an SP-owned path when promoted.
Next steps¶
Deploy framework from local machine — deploy the framework from your laptop (dev inner loop)
Deploy pipeline bundles from local machine — deploy a pipeline bundle from your laptop
Setting up CI/CD — automate deploy with CI/CD, including framework versioning