search

aws-samples / eks-workshop

AWS Workshop for Learning EKS
https://eksworkshop.com
MIT No Attribution
804 stars 1.24k forks source link

Feedbacks from kubeflow workshop #620

Closed Jeffwan closed 1 year ago

Jeffwan commented 4 years ago

Problems

  1. Attendees copy this export AWS_REGION= and have errors in https://eksworkshop.com/030_eksctl/test/

image https://eksworkshop.com/020_prerequisites/workspaceiam/

People may do themselves and miss what instructor are talking.

  1. People still go to setup SSH key and can not find next section https://eksworkshop.com/020_prerequisites/sshkey/

  2. Choose right image - for fairing and JupyterNotebook. People still use default one for fairing.. I notice someone use GPU container and failed to import tensorflow because container doesn't have GPU.

  3. Fairing - kubeflow-pipeline-data is inconspicuous. image Users actually have to change to their own bucket. We want to change the code make sure if use env variables by default.

  4. Can not submit Sagemaker jobs. Very similar to this. https://github.com/aws-samples/eks-workshop/issues/603 The reason is users may miss step to use sagemaker user. I would suggest to use IAM user for all workshops. Create a workshop user and attach Sagemaker, S3, ECR permissions.

  5. Batch Transformation Failure https://github.com/aws-samples/eks-workshop/issues/521

  6. Increase cluster size is probably unnecessary image https://eksworkshop.com/advanced/420_kubeflow/install/ 3 nodes probably are good enough. This takes extra 3 mins.

  7. Move Training & Serving in Notebook Users said model training and model inference are not part of Kubeflow. We probably want to consider to move to jupyter notebook based training and inference. Just an example to practice notebook.

  8. Shorten time to setup cluster. If people don't care the cluster setups, probably write a scripts to setup users and bring up a cluster

  9. Users copy & paste output from instruction. I would suggest no to use exact same markdown for outputs. image https://eksworkshop.com/advanced/420_kubeflow/pipelines/ For example, this is not something we want user to copy and run.

Jeffwan commented 4 years ago

@arun-gupta @dalbhanj

github-actions[bot] commented 1 year ago

This issue is stale because it has been open for 30 days with no activity.

github-actions[bot] commented 1 year ago

This issue was closed because it has been inactive for 14 days since being marked as stale.