Note: this is work-in-progress
Import manifests from seldon-server (need to have it checked-out in this dir):
./scripts/import-infra
git add infra
git commit -m 'Add the Seldon stack to my GKE cluster!'
git push
Run setup finaliser script:
./scripts/finilise-setup
Add the Deep MNIST jobs:
./scripts/import-app seldon-server/kubernetes/conf/examples/tensorflow_deep_mnist/train-tensorflow-deep-mnist.json
./scripts/import-app seldon-server/kubernetes/conf/examples/tensorflow_deep_mnist/load-model-tensorflow-deep-mnist.json
git add apps
git commit -m 'Deploy some app workloads!'
git push
kubectl get jobs -w -o wide
Create Seldon client:
./scripts/seldon-cli client --action setup --db-name ClientDB --client-name deep_mnist_client
Deploy the example TensorFlow app:
./scripts/start-microservice --type prediction --client deep_mnist_client -p tensorflow-deep-mnist /seldon-data/seldon-models/tensorflow_deep_mnist/1/ rest 1.0
Check service status:
kubectl get service,deployment tensorflow-deep-mnist
Start the front-end app:
./scripts/start-front-end
This just created a deployment and a service in a very ad-hoc style, clearly we want this production-ready with GitOps!
We can export & check manifests like this:
kubectl get deployment/tensorflow-deep-mnist --output=yaml --export=true > apps/tensorflow-deep-mnist.Deployment.yaml
kubectl get service/tensorflow-deep-mnist --output=yaml --export=true > apps/tensorflow-deep-mnist.Service.yaml
kubectl get deployment/deep-mnist-webapp --output=yaml --export=true > apps/deep-mnist-webapp.Deployment.yaml
kubectl get service/deep-mnist-webapp --output=yaml --export=true > apps/deep-mnist-webapp.Service.yaml
Now lets's review this, we certainly need to adjust the service type...
git add apps
git commit -m 'gitOps := "AWESOME"'