Would like to check with you, is the Object Cloud Storage on IBM cloud an option to store the flow? I don’t see it in the docs, just wondering. Thanks.

Hey @Zheng Xie, it is not at the moment, bit I think we’d accept a contribution for this if you happen to be interested? Even if you write your own, you need the edited code on the agent side to pull the flow. Do you think maybe it can be done with webhook storage?

Hi Kevin, thanks. I have seen your video and very happy to get your help here 🙂 the infrastructure I run the python code that takes training data and train the models currently is my laptop. I would like it to be open to other users. They can upload their new training data to a cloud location, then trigger the flow. so I have to move the flow and agent out of my laptop. I have subscription to IBM cloud, a public kubernetes cluster and other services. Thinking of doing the following: Refactor the model training code to be prefect flows and tasks, Packaging it into docker image and deploy to the Kubernetes cluster. The code takes in training data from a cloud storage, like Box. Will this be ok? What is a good option for the prefect flow storage?

Thanks for watching! That makes sense. You can use Docker storage since it works well with KubernetesRun and you can package you dependencies in that image. That way, the loading of data is just an API call inside a Task and it’s not Prefect responsible for it. This setup makes perfect sense.

Hi Kevin, do you have any example by chance on deploying the flow and tasks in a generic Kubernetes cluster on a cloud provider? I found this Deploy Code to AWS Kubernetes with Prefect: a Step-by-Step ... but it isn’t written in a way I can adapt it to IBM Cloud. I am not a computer wizard…

I think the easiest is that you have your agent in the same cluster so that it can deploy flows there. This command will create the deployment for the agent for you. You can then use the

kubectl apply

to get the agent. Once you have the agent online, just use

KubernetesRun

and then the flow gets deployed on the cluster.