Once you are finished using a dataset, you may wish to remove a particular version of a datastore.
Please be careful!
Datastore deletion is immediate and unrecoverable. If the delete operation reports that the request was successful (e.g. the datastore is not in use by a Session or Run when the delete request was issued), then the data has already begun a rapid removal routine and there is nothing that can cancel, interrupt, or reverse the process!
Deleting a datastore can be done via either the CLI or the UI, and requires that you specify both the Datastore name and desired version to delete.
Multiple datasets uploaded to a Datastore using the same name will create independent versions. Each dataset / Datastore instance is automatically assigned a unique (and previously unused) version number starting at 1 and incrementing on each subsequent upload.
If you only need some specific version of a datastore name, it is SAFE to delete any/all other versions of the Datastore. The data contained in each version is complete, immutable, and independent of any other Datastore version.
Using the CLI
Let's say I created a few different variations of the same Datastore named
"my-training-data" as I was experimenting with pre-processing algorithms for a model I was
in the process of training. I have versions
4 available, but I'm
only happy with versions
4. Because I'm a conciencious user who doesn't want
to abuse the free storage I'm using for this data, I'm going to delete versions
3. I look at the
grid datastore delete command reference and
am able to execute the following commands to clean up after myself!
grid datastore delete --name my-training-data --version 1
grid datastore delete --name my-training-data --version 3
Using the UI
Deleting a single version of a datastore in the UI is currenly not supported. If you try to delete a datastore with multiple versions, it will delete all versions of that datastore.