ai-verify for kubernetes

Docker Image Changes

The default docker image needs to be modified in order to run in kuberenetes.

We copy over the new run.sh to accomplish the following

Remove ssl arguments from the ng serve command
add the --disable-host-check argument to the ng serve command

Run the following commands to build and push the correct docker image:

cd docker
docker build -t <repo-name>/<image-name>:latest .
docker push <repo-or-user-name>/<image-name>:latest

Kubernetes Manifests

The kubernetes manifests all live in the kubernetes directory.

After you have modified the appropriate files below, deploy by running:

cd kubernetes
kubectl apply -k .

We will use a kubectl tool called kustomize to deploy the files. There are the following components that make up the app deployment.

patch-deployment.yaml <-Modify Me

The only line you should have to change here is the image line. Make sure it points to the image you built and push in the steps above

patch-ingress.yaml <- Modify Me

Make sure to modify the url (hosts) lines. This will need to have a url that the ingress controller on the kubernetes cluster is configured to resolve. If your cluster leverages cert-manager or some other cert management system make sure the annotations and tls sections are configured appropriately, otherwise you will most likely need to remove those sections.

Do not modify any remaing files

kustomization.yaml

This will tell kubectl where the core manifests are as well as where the patch files live

namespace.yaml

This simply creates the namespace where the deployment and all other components will live.

deployment.yaml

This is where we define important pod information such as what container image to use and any ports the image listen in on.

service.yaml

This will create the connective tissue between the ingress object (dns entry) and the main deployment pod

ingress.yaml

This is where we define what url we will use as well as any information about generating tls certs for the url

Other notes

I generated all the default manifest files as a starting place running the following commands:

kubectl create namespace ai-verify --dry-run=client --output=yaml > namespace.yaml
kubectl create deployment ai-verify --image=harbor.web.craigcloud.io/cnvrg/ai-verify-custom:latest --port=4200 --dry-run=client --output=yaml --namespace=ai-verify > deployment.yaml
kubectl expose deployment ai-verify --type=ClusterIP --port=4200 --name=ai-verify --namespace=ai-verify --dry-run=client --output=yaml > service.yaml
kubectl create ingress ai-verify --class=contour --rule="ai-verify.web.craigcloud.io/*=ai-verify:4200,tls=ai-verify" --annotation="cert-manager.io/cluster-issuer=letsencrypt-prod" --annotation="ingress.kubernetes.io/force-ssl-redirect=true" --annotation="kubernetes.io/tls-acme=true" --dry-run=client -n ai-verify -o yaml > ingress.yaml

I noticed that this container is running:

redis
mongodb
ai-verify

It is basically a full app unto itself, this is a good reason why cnvrg would not be a great target to run this. It really should be broken apart into three different kubernetes deployments (redis/mongo/ai-verify). This would allow you to create persistence for the mongo and redis pieces to allow ai-verify to die and restart without losing any progress!

nctiggy/ai-verify