The following are the setup steps required.
The following are the prerequisites:
- Have completed the Pre-Boot Camp Learning as suggested to you.
- Have a Cisco provided Mac laptop with all recommended software, especially VPN. You will be primarily working on your laptops and connecting to a kubernetes (k8s) cluster runnin on Google Kubernetes Engine.
- Have a gmail address.
- Each team will be given a service account.
- Install gcloud on your macbooks.
- Install kubectl on your macbooks.
- Activate service account (
service-acc-nameandjson-file-namewill be provided)
gcloud auth activate-service-account <service-acc-name> --key-file=<json-file-name> --project=ml-bootcamp-2018
- Get kubeconfig for your cluster (
cluster-nameandzone-namewill be provided)
gcloud container clusters get-credentials <cluster-name> --zone <zone-name>
- Enable admin cluster role binding (
your-user-cluster-admin-bindingwas retrieved in the previous step)
kubectl create clusterrolebinding your-user-cluster-admin-binding --clusterrole=cluster-admin --user=<service-acc-name>
Ensure that the MNIST example in the KFLab repo is accessible. Read through the README.md in that repo so that you understand how the application is structured.
NOTE: The example code is for running on GKE. Should not matter much since all you need is a k8s cluster but it is important to note that the bootcamp will be running on an in-house k8s cluster.
-
Kubernetes. A nice overview can be found at https://jvns.ca/blog/2017/06/04/learning-about-kubernetes/. Here are the slides from the presentation.
-
Kubeflow. An overview of Kubeflow can be found at the official Kubeflow documentation. Here are the slides from the presentation.
This section is for the administrators of the bootcamp only. Information here is not relevant to others and links from here will usually not be available to anyone but administrators.
All admin related scripts are kept at https://github.com/CiscoAI/bootcamp-ops.