BentoML

~You might have better luck over on the BentoML slack.~

My bad, for some reason I thought I was over at the DataScienceClub slack...

image.png

Hi, can you show me the image builder pod logs?

You can use `kubectl`:

```kubectl -n yatai-builders logs -f yatai-bento-image-builder-iris-classifier-blm-246-tfgwofxlg-6```
Or on the UI:

so running the kubectl command gives me:
```Defaulted container "builder" out of: builder, bento-downloader (init), model-downloader-0 (init)
Error from server (BadRequest): container "builder" in pod "yatai-bento-image-builder-iris-classifier-blm-246-tfgwofxlg-6" is waiting to start: PodInitializing```
Looking at UI, I get the following results

Can you show your helm release version of yatai-deployment?

```helm -n yatai-deployment list```

huhh, that’s weird. I’m getting: `Error: Kubernetes cluster unreachable: Get "<https://127.0.0.1:61729/version>": net/http: TLS handshake timeout`.Could this in anyway be related to this setting in my docker?

Do you use minikube? minikube is not related on the DockerDesktop's k8s

You can use `minikube status` to check it

Yes I’m using minikube to run this. Running minikube status gives me:
```type: Control Plane
host: Running
kubelet: Running
apiserver: Running
kubeconfig: Configured```

How about this command?
```kubectl get node -o wide```

Your minikube looks to be working correctly.

I just found this bug and fixed it and released this fix, you can use this command to upgrade `yatai-deployment`

```helm repo update bentoml
helm upgrade yatai-deployment bentoml/yatai-deployment -n yatai-deployment```

~do i need to restart yatai for this to start working?~

I have 1 final question on this, the URL endpoint given here does not seem to load. Do i need to do anything extra to access this model?

You should use `hyperkit` driver to prevent the network limitation of docker on macOS

<https://github.com/bentoml/Yatai#%EF%B8%8F-prerequisites>