Kfserving: No Activitiy related to knative-service on 1.18 of K8's

[bharath0208]

Hi Team,

I have deployed the https://raw.githubusercontent.com/kubeflow/manifests/master/kfdef/kfctl_istio_dex.yaml on my Kubernetes(v1.18.4). We have tried running some examples like sklearn and hello-world. We dont see any pods running on admin namespace.

Additional Info:

kubectl describe inferenceservice -n admin
Name:         sklearn-iris
Namespace:    admin
Labels:       <none>
Annotations:  kubectl.kubernetes.io/last-applied-configuration:
                {"apiVersion":"serving.kubeflow.org/v1alpha2","kind":"InferenceService","metadata":{"annotations":{"sidecar.istio.io/inject":"false"},"nam...
              sidecar.istio.io/inject: false
API Version:  serving.kubeflow.org/v1alpha2
Kind:         InferenceService
Metadata:
  Creation Timestamp:  2020-07-17T18:23:45Z
  Generation:          1
  Managed Fields:
    API Version:     serving.kubeflow.org/v1alpha2
    Manager:         kubectl
    Operation:       Update
    Time:            2020-07-17T18:23:44Z
  Resource Version:  366999
  Self Link:         /apis/serving.kubeflow.org/v1alpha2/namespaces/admin/inferenceservices/sklearn-iris
  UID:               c8d308b4-639d-4cb7-ad11-4f26231a2329
Spec:
  Default:
    Predictor:
      Sklearn:
        Resources:
          Limits:
            Cpu:     1
            Memory:  2Gi
          Requests:
            Cpu:          1
            Memory:       2Gi
        Runtime Version:  v0.3.0
        Storage Uri:      gs://kfserving-samples/models/sklearn/iris
Events:                   <none>

We see that there is no activity related to knative-service on the cluster. Is there a work-around for this problem.

[issue-label-bot[bot]]

Issue-Label Bot is automatically applying the labels:

| Label | Probability |
| ------------- | ------------- |
| area/inference | 0.79 |

Please mark this comment with :thumbsup: or :thumbsdown: to give our bot feedback!
Links: app homepage, dashboard and code for this bot.

[yuzisun]

@bharath0208 There is known issue that KFServing does not work with k8s 1.18, if you can downgrade to 1.17 that should work.

[bharath0208]

@yuzisun any idea on when of kfserving will be supporting version 1.18 of k8's

[issue-label-bot[bot]]

Issue-Label Bot is automatically applying the labels:

| Label | Probability |
| ------------- | ------------- |
| area/engprod | 0.54 |

Please mark this comment with :thumbsup: or :thumbsdown: to give our bot feedback!
Links: app homepage, dashboard and code for this bot.

[yuzisun]

@bharath0208 there is an upstream issue https://github.com/kubernetes-sigs/controller-tools/issues/438 for the tool KFServing uses which generates the invalid inference service crd on k8s 1.18

[Bhavesh-Joshi]

It's a long time now. Any idea on when of kfserving will be supporting version 1.18 of k8's ????

[CreatorSky]

@yuzisun I see there is a PR (https://github.com/kubeflow/kfserving/pull/1128) to enable support for k8s 1.18, can we apply the update (0.4.1) to existing kfserving (under kubeflow namespace) installation?

[yuzisun]

@CreatorSky @Bhavesh-Joshi we are patching 1.18 support to KFS 0.4.1 release and the fix is going to be in kubeflow 1.2 release.

[yuzisun]

released in 0.4.1

[ontheway16]

@bharath0208 There is known issue that KFServing does not work with k8s 1.18, if you can downgrade to 1.17 that should work.

@yuzisun , Is this information still valid ? Should we install 1.17? (I am about to setup a new local cluster, looking to find best working versions of k8s, kubeflow, kfserving, istio, knative, docker etc..)
Thanks.

[chinmay22897]

@bharath0208 There is known issue that KFServing does not work with k8s 1.18, if you can downgrade to 1.17 that should work.

@yuzisun , Is this information still valid ? Should we install 1.17? (I am about to setup a new local cluster, looking to find best working versions of k8s, kubeflow, kfserving, istio, knative, docker etc..)
Thanks.

@ontheway16 Did you find any good versions maping so that every thing works for local cluster at least? I am using v0.7.0 of kubeflow as of now as other newer versions are not working on my minikube setup (for which I have already raised issues but looks like there is no support for them as of now)

[ontheway16]

@chinmay22897 hi, I am using regular kubernetes install, but since I am focusing to inference only, at the moment, I decided to follow standalone KFserving install route. All seems working with 1.17, but I continue to test, still learning.