No information on deleting datasets

[davystrong]

I created a dataset with the wrong type. I want to recreate it with the same name but the correct type. However, I can't find any information on how to delete a dataset. The linked docs page mentions archiving but not how to delete datasets. I feel this should be documented somewhere.

---

Document Details

⚠ Do not edit this section. It is required for learn.microsoft.com ➟ GitHub issue linking.

• ID: cf5a19a0-c37e-4567-0c30-7efe1c433dfb
• Version Independent ID: 8c4e79d0-0f84-3a28-bfb4-9f972e02a37f
• Content: az ml data
• Content Source: latest/docs-ref-autogen/ml/data.yml
• GitHub Login: @rloutlaw
• Microsoft Alias: routlaw

[yonzhan]

Thank you for opening this issue, we will look into it.

[SwathiDhanwada-MSFT]

@davystrong Apologies for the late response. You can unregister data assets and archive jobs, but these operations don't delete the data. To entirely remove the data, data assets and job data require deletion at the storage level. Storage level deletion happens in the portal.

As an initial step, traverse to ML Studio -> Select Data under Assets -> Select the Assets you wish to delete -> Click on the Assest

The asset contains information as shown below. Click on View in Azure portal browse as highlighted in the image.

View in Azure portal redirects to the path where the data asset is located in storage account. Click on Delete. This deletes the asset.

As per this document, its mentioned that you can export or delete your workspace data using either the portal graphical interface or the Python SDK.

If you want to do via Azure CLI, its bit of complicated process.

• First, you need to identify the datastore using the path where the data asset is located. You can use az ml data show command.
• Using datastore , identify the account name and container name where data assets is stored. You can use az ml datastore show command.
• Traverse to the storage account and identify the container where data assets are stored. The exact path can be identified from the first step. You can use az storage blob delete command to delete the data asset.

The above information is shared based on the assumption that you are using version 1 API to create the data asset. I assume that its same process for version 2 as well. Do let me know if that does not work for you or if you have further questions.

[davystrong]

Hi @SwathiDhanwada-MSFT, thanks for getting back. What I'm asking about isn't deleting the underlying data, it's the very first part of your answer: deleting the data asset "reference" in ML Studio. There seems to be no documented way of doing this for v2 data assets. For v1 (file and tabular) there was the option to unlink the asset, but for v2 this has been replaced with an archive button.

[SwathiDhanwada-MSFT]

@davystrong I am bit confused about your request. The Archive button does what you are asking for, it removes the link between the data asset and underlying storage.