GitOps for Machine Learning

TBD

Prerequisites

1. Package installation

Install Minikube
Install Helm
Install Argo-CD CLI
Install Argo-Workflows CLI

You can install the all prerequites with Homebrew:

brew install minikube helm argocd argo

2. Fork and clone this repository

TBD

3. Register a SSH key

Create a new SSH key with ECDSA encryption and add it to GitHub.

ssh-keygen -t ecdsa -b 521 -C "your@email-address.com"
cat ~/.ssh/id_ecdsa.pub  # show the public key created

Cluster setup

1. Kubernetes cluster initialization

make cluster # k8s cluster setup and argo-cd installation
make tunnel  # need to be active for the cluster's LoadBalancer access

2. Login ArgoCD and Add the repository

Check all argo-cd pods are in Running status.

kubectl get pods

# NAME                                                       READY   STATUS    RESTARTS   AGE
# argo-cd-argocd-application-controller-0                    1/1     Running   0          22m
# argo-cd-argocd-applicationset-controller-ccb4b5fc5-l2nwj   1/1     Running   0          22m
# argo-cd-argocd-dex-server-658b5c88f4-qzdl7                 1/1     Running   0          22m
# argo-cd-argocd-notifications-controller-67dcf4c8-dt28f     1/1     Running   0          22m
# argo-cd-argocd-redis-6d4576dbfb-77s6m                      1/1     Running   0          22m
# argo-cd-argocd-repo-server-744744d646-dx5zx                1/1     Running   0          22m
# argo-cd-argocd-server-f66f985c-xz5hl                       1/1     Running   0          22m

make init-argocd

# ARGOCD_PW= &&\
#                 argocd login localhost:8080 --username admin --password ******** --insecure
# 'admin:login' logged in successfully
# Context 'localhost:8080' updated
# argocd repo add git@github.com:Curt-Park/gitops-for-machine-learning.git --ssh-private-key-path ~/.ssh/id_ecdsa
# Repository 'git@github.com:Curt-Park/gitops-for-machine-learning.git' added

3. Install other apps

Create applications on Argo-CD. Argo-CD will watch the git remote branch and synchronize the applications with it if it changes.

make argo-cd
make argo-workflows
make minio

You can see the login information as follows.

for Argo-CD:

# user: admin
# password:
kubectl get secret argocd-initial-admin-secret -o jsonpath="{.data.password}" | base64 --decode

for MinIO:

# AccessKey
kubectl get secret minio -o jsonpath='{.data.accesskey}' | base64 --decode

# SecretKey
kubectl get secret minio -o jsonpath='{.data.secretkey}' | base64 --decode

4. Create a rolebinding

In order for Argo to support features such as artifacts, outputs, access to secrets, etc. it needs to communicate with Kubernetes resources using the Kubernetes API. To communicate with the Kubernetes API, Argo uses a ServiceAccount to authenticate itself to the Kubernetes API. By default, it runs workflows with the default ServiceAccount. Here, we will grant the default ServiceAccount admin privileges.

kubectl create rolebinding default-admin --clusterrole=admin --serviceaccount=default:default

Workflow Submission

Hello World

argo submit https://raw.githubusercontent.com/argoproj/argo-workflows/master/examples/hello-world.yaml --watch

# How it looks like:
# wget https://raw.githubusercontent.com/argoproj/argo-workflows/master/examples/hello-world.yaml
# cat hello-world.yaml
#
## apiVersion: argoproj.io/v1alpha1
## kind: Workflow
## metadata:
##   generateName: hello-world-
##   labels:
##     workflows.argoproj.io/archive-strategy: "false"
##   annotations:
##     workflows.argoproj.io/description: |
##       This is a simple hello world example.
##       You can also run it in Python: https://couler-proj.github.io/couler/examples/#hello-world
## spec:
##   entrypoint: whalesay
##   templates:
##   - name: whalesay
##     container:
##       image: docker/whalesay:latest
##       command: [cowsay]
##       args: ["hello world"]

Argo-Workflows will show how it goes.

The container logs:

 _____________
< hello world >
 -------------
    \
     \
      \
                    ##        .
              ## ## ##       ==
           ## ## ## ##      ===
       /""""""""""""""""___/ ===
  ~~~ {~~ ~~~~ ~~~ ~~~~ ~~ ~ /  ===- ~~~
       \______ o          __/
        \    \        __/
          \____\______/

Cluster shutdown

make finalize

Curt-Park / gitops-for-machine-learning