Artifact Registry
This feature is in Beta. To use it, a workspace admin must turn on Databricks Artifact Registry from the Previews page. See Manage Databricks previews. Databricks will announce pricing for Artifact Registry when it becomes generally available.
Artifact Registry stores container images as governed assets in Unity Catalog. Use it to push images with standard Docker commands, control access with Unity Catalog privileges, and make images available to supported Databricks workloads.
Key concepts
- Artifact: A named container image stored in Unity Catalog, identified by a three-level name in the format
catalog.schema.artifact. - Tag: A mutable alias that points to an artifact version using the format
catalog.schema.artifact:tag. Pushing a tag again can change the version that the tag resolves to. For repeatable deployments, avoid mutable tags such aslatestand reference a digest when the workload accepts digest references. - Image digest: An immutable, content-addressed identifier for a specific version of an artifact.
- Image acceleration: An optimization that Databricks applies automatically after you push an eligible image to Artifact Registry. This process converts the image to a format designed to reduce pull latency for supported workloads. To check an artifact version's image acceleration status, see Check image acceleration status.
How Artifact Registry works
Publishers and users work with container images in Artifact Registry through the following lifecycle:
- Build and push: A publisher builds an image, tags it with a three-level Unity Catalog name in the format
catalog.schema.artifact:tag, and pushes it to the workspace registry. See Artifact name requirements and Push an image. - Govern: Artifact Registry stores the image under the specified Unity Catalog catalog and schema. Unity Catalog privileges control who can publish, discover, and read the image. See Grant permissions.
- Discover: Users call the Software Artifacts API to discover available images and versions and to check image acceleration status. See Discover images and versions.
- Use: Users reference an image when they create supported Databricks workloads, such as jobs that run custom container images.
Limitations
The following limitations apply:
-
Availability:
-
Artifact Registry is available in the following regions:
asia-south1asia-southeast1europe-west1europe-west2europe-west3us-central1us-east1us-east4us-west1
-
-
Supported use:
- You can use images in Artifact Registry only with supported Databricks workloads.
- You cannot use Docker or another registry client to pull these images or download image blobs directly.
-
Authentication:
- The workspace OAuth access token used for Docker authentication expires after one hour. A push that takes longer can fail authorization for a layer uploaded after expiration. Retry the push so that the Docker credential helper requests a new token.
-
Image acceleration:
- Images larger than 30 GB cannot be converted for image acceleration.
- Image acceleration does not expose a failed conversion state. See Check image acceleration status.
-
Discovery and management:
- Catalog Explorer does not provide an interface for images.
- The registry endpoint does not support Open Container Initiative (OCI) tag-list requests or delete requests.
- The Software Artifacts API returns at most 10 tags for a version when more than 10 tags point to the same digest.
-
Artifact names:
- Artifact names have character and formatting restrictions. See Artifact name requirements.
Additional resources
- To configure authentication, grant access, push an image, and discover its versions, see Get started with Artifact Registry.
- To resolve common Beta access, authentication, push, discovery, and image acceleration issues, see Troubleshoot Artifact Registry.