Skip to content

docs: document storageContainerName for explicit ClusterStorageContainer selection #693

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
RunlinXiao wants to merge 1 commit into
base: main
Choose a base branch
from
+486 −14
Open

docs: document storageContainerName for explicit ClusterStorageContainer selection #693

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
37 changes: 36 additions & 1 deletion docs/model-serving/storage/storage-containers/storage-containers.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,7 +46,7 @@ In a _ClusterStorageContainer_ spec, you can specify credentials for cloud stora

:::warning

If a storage URI is supported by two or more _ClusterStorageContainer_ CRs, there is no guarantee which one will be used. **Please make sure that the URI format is only supported by one ClusterStorageContainer CR**.
If a storage URI is supported by two or more _ClusterStorageContainer_ CRs and `storageContainerName` is not set, there is no guarantee which one will be used. Set [`spec.predictor.storageContainerName`](#explicitly-selecting-a-storage-container) on the `InferenceService` to explicitly select the one to use.

:::

Expand Down Expand Up @@ -162,6 +162,41 @@ The respective secret should be created in the same namespace as the `InferenceS
kubectl create secret generic hf-secret --from-literal=HF_TOKEN=<your_huggingface_token>
```

## Explicitly Selecting a Storage Container

By default, KServe selects a `ClusterStorageContainer` by matching the storage URI against each CR's `supportedUriFormats`. When more than one CR supports the same URI format, you can set `storageContainerName` on the predictor spec to explicitly select which one to use, similar to `storageClassName` on a PersistentVolumeClaim.

For example, both the `default` and the `hf-hub` ClusterStorageContainers above support the `hf://` URI format. To make sure the `hf-hub` container (which carries the Hugging Face credentials) is used, reference it by name:

```yaml
apiVersion: serving.kserve.io/v1beta1
kind: InferenceService
metadata:
name: huggingface-llama3
spec:
predictor:
storageContainerName: hf-hub
model:
modelFormat:
name: huggingface
storageUri: hf://meta-llama/meta-llama-3-8b-instruct
```

When `storageContainerName` is set, there is no fallback to auto-matching. The referenced `ClusterStorageContainer` must:

- exist in the cluster,
- not be disabled,
- have `workloadType: initContainer` (the default),
- support the model's `storageUri` in its `supportedUriFormats`.

If any of these checks fail, the deployment fails with an error instead of falling back to another storage container.

:::note

`storageContainerName` is only available on the predictor spec. Explainers and transformers always use URI-based auto-matching, and `LLMInferenceService` does not support this field yet.

:::

## Spec Attributes

Spec attributes are in [API Reference](../../../reference/crd-api.mdx#clusterstoragecontainer) doc.
Loading