Can we attach pvc to model-initializer(storage-initializer) ?

[hemantha-kumara]

Hi,
I am trying to run XGBOOST_SERVER serving by using model present in pv

apiVersion: machinelearning.seldon.io/v1alpha2
kind: SeldonDeployment
metadata:
  name: xgboost
spec:
  name: iris
  predictors:
  - graph:
      children: []
      implementation: XGBOOST_SERVER
      modelUri: file://seldon-models/xgboost/iris
      name: classifier
    name: default
    replicas: 1 

but problem is I am not able to attach pvc to model-initializer and hence model is not accesible to storage-initializer.
Due to mentioned problem i am not able to serve models.

can anyone pls help me to attach pvc to init container?

[ryandawsonuk]

The format is pvc://. It proved tricky to get working with tensorflow in a recent issue but we did manage it.

[hemantha-kumara]

Thanks @ryandawsonuk
pvc://<claim name> solving initcontainer pvc issue.

[hemantha-kumara]

I am trying serving in on-prem airgap deployments where docker.io is not accessible,
as DefaultXGBoostServerImageNameRest is hardcoded to use docker.io, now serving pod is going to ImagePullBackOff error.

[ryandawsonuk]

@hemantha-kumara yeah we want to make those images configurable and are working on the change.

[hemantha-kumara]

Thanks @ryandawsonuk
I was expecting #1118 changes and thanks again for pointing fix.