How-to: resolve a network volume by name and attach it to a serverless endpoint
When you attach a network volume to a serverless endpoint you do not need to
look the volume's id up yourself. rp serverless create accepts the volume by
name and resolves it internally, then pins the endpoint to that volume's
datacentre. A pod is different: rp pod create takes a volume only by id —
that path is covered in
Create a network volume and attach it to a pod by name.
Steps
- Attach a network volume to a serverless endpoint by name. Pass the volume
name to
--network-volume <name>; the CLI resolves it to an id and scopes the endpoint to the volume's datacentre for you:
$ rp serverless create --name my-endpoint --template tmpl_abc \
--network-volume hf-cache --gpus-from-volume hf-cache \
--workers-min 0 --workers-max 3 --idle 600
--gpus-from-volume here picks in-stock serverless GPU types from a fixed
four-type preference list — account-wide stock, not filtered by the
volume's datacentre; the pinning comes entirely from --network-volume
(see Scope a serverless endpoint to a network volume's
datacentre).
- (Optional) Attach by id instead, when you already hold one. The by-id
equivalents are
--network-volume-id <id>(one) and--network-volume-ids <id,id>(several, template path only):
$ rp serverless create --name my-endpoint --template tmpl_abc \
--network-volume-id vol_abc123 --gpu "NVIDIA L4" \
--workers-min 0 --workers-max 3 --idle 600
To read a volume's id or datacentre from its name in one line:
$ rp volume list --json --jq '.[] | select(.name=="hf-cache") | .id'
$ rp volume list --json --jq '.[] | select(.name=="hf-cache") | .dataCenter'
Notes
- The serverless path resolves the name automatically:
--network-volume <name>is the by-name flag, and the by-id flags are there for when you already know the id. - The endpoint lives in the same datacentre as the volume — enforced for you,
since the datacentre is pinned from the volume; there is no separate
--dcstep to perform. - Only some datacentres expose the S3 API. A volume created in one that does
not still attaches to an endpoint, but
rp volume sync/rp volume lswill not be able to reach it —rp stock dcmarks the ones that do.