Storage Mounts
Similar to ML Jobs, Pipelines enable customers to mount File Storage and Object Storage as local folders for pipeline steps.
This simplifies the process of accessing data stored in File or Object Storage, removing the need for installing extra libraries or writing specialized code. Using standard OSs or language libraries, users can access shared file systems and object storage buckets in the same way they access local files and folders.
You can also override storage mounts in pipeline runs, enabling one-off changes for a specific run. This is essential for SaaS customers who use pipelines within ML Applications and need to override storage mounts for a particular SaaS customer (tenant).
Configuring Storage Mounts
Ensure you have the appropriate Policies applied.
Storage mounts can be configured at two levels:
- Pipeline level: The storage mount configuration is applied to all steps, unless overridden.
- Step level: The storage mount configuration is applied only to the specific step.
Storage mounts apply only to the following step types:
- Script step type. Steps defined by a script.
- BYOC step type. Steps defined by a container image.
The configuration is managed in the Storage Mounts configuration panel.
You can configure no more than five storage mounts for a pipeline or a step.
Storage Mount Overrides
You can specify storage mount overrides for Pipeline Runs the same way as for Pipelines. For details about storage mounts, see Storage Mounts.
When creating a pipeline run, you can define the override at the pipeline level or for a pipeline step.