Skip to content

local-deployment-guide.md

Latest commit

 

History

History
132 lines (86 loc) · 4.17 KB

local-deployment-guide.md

File metadata and controls

132 lines (86 loc) · 4.17 KB

Local Deployment Guide

Local kubernetes deployment of Model Registry

To test the BFF locally without mocking the k8s calls the Model Registry backend can be deployed locally using kind.

Prerequisites

The following tools need to be installed in your local environment:

Note: all of the above tools can be installed using your OS package manager, this is the preferred method.

Setup

1. Create a kind cluster

Create a local cluster for running the MR backend using the following command:

kind create cluster

Kind will start creating a new local cluster for you to deploy, once it has completed verify you can access the cluster using kubectl by running:

kubectl cluster-info

If everything is working correctly you should see output similar to:

Kubernetes control plane is running at https://127.0.0.1:58635
CoreDNS is running at https://127.0.0.1:58635/api/v1/namespaces/kube-system/services/kube-dns:dns/proxy

2. Create kubeflow namespace and switch to it

Create a namespace for model registry to run in, by default this is kubeflow, run:

kubectl create namespace kubeflow

then, switch to that kubeflow namespace

  kubectl config set-context --current --namespace=kubeflow

3. Deploy Model Registry to cluster

You can now deploy the MR backend to your newly created cluster using the kustomize configs in the MR repository by running:

kubectl apply -k "https://github.com/kubeflow/model-registry/manifests/kustomize/overlays/db"

Wait for the model registry deployment to spin up, alternatively run:

kubectl wait --for=condition=available -n kubeflow deployment/model-registry-deployment --timeout=1m

This command will return when the cluster is ready. To verify this now run:

kubectl get pods -n kubeflow

Two pods should be listed, model-registry-db-xxx and model-registry-deployment-yyy both should have a status of Running.

NOTE: Issues running on arm64 architecture

There is currently an issue deploying to an arm64 device such as a Mac with an M-series chip. This is because the MySql image tag deployed by the manifests does not have an arm64 compatible image. To work around this you can use a modified manifest deployed in a fork of the repo - you can use this by running the below command instead of the first command in section 3 of this guide.

kubectl apply -k "https://github.com/alexcreasy/model-registry/manifests/kustomize/overlays/db?ref=kind"

Note: an issue has been filed regarding this ticket here:

4. Setup a port forward to the service

In order to access the MR REST API locally you need to forward a local port to 8080 on the MR service. Run the following command: (Note: this command starts a long-running service and will not exit the shell)

kubectl port-forward svc/model-registry-service -n kubeflow 8080:8080

Note: you can change the local forwarded port by changing the first port value, e.g. 4000:8080 will forward port 4000 to the MR service.

5. Test the service

In a separate terminal window to the previous step, test the service by querying one of the rest endpoints, for example:

curl http://localhost:8080/api/model_registry/v1alpha3/registered_models

You should receive a 200 response if everything is working correctly, the body should look like:

{"items":[],"nextPageToken":"","pageSize":0,"size":0}

6. (Optional) Run BFF locally in Dev Mode

To access your local kind cluster when running the BFF locally, you can use the DEV_MODE option. This is useful for when you want to test live changes on real cluster. To do so, simply run:

make run DEV_MODE=true

You can also specify the port you are forwarding to if it is something other than 8080:

make run DEV_MODE=true DEV_MODE_MODEL_REGISTRY_PORT=8081