feat: migration script for V1 to V2 components

README.md

@@ -6,7 +6,7 @@
 It exposes the access controlled web interfaces for Kubeflow components and more.
 
 > ⚠️ __Note__ ⚠️
-> 
+>
 > We are currently moving the Kubeflow Dashboard codebase from [`kubeflow/kubeflow`](https://github.com/kubeflow/kubeflow) to this repository ([`kubeflow/dashboard`](https://github.com/kubeflow/dashboard)).
 > Please see [`kubeflow/kubeflow#7549`](https://github.com/kubeflow/kubeflow/issues/7549) for more information.
 
@@ -38,6 +38,34 @@ Please refer to the [Installing Kubeflow](https://www.kubeflow.org/docs/started/
 
 The official documentation for Kubeflow Dashboard can be found [here](https://www.kubeflow.org/docs/components/central-dash/).
 
+## Migrate from V1
+
+In the past, up until the Kubeflow 1.10 release, the components hosed in this repository were living in the
+[`kubeflow/kubeflow`](https://github.com/kubeflow/kubeflow) repository.
+
+To accomodate the migration, the first `v2.0.0` release of this repository contains the same artifacts
+from the corresponding [`v1.10.0`](https://github.com/kubeflow/kubeflow/releases/tag/v1.10.0) tag from the `kubeflow/kubeflow` repo. This was done to ensure a smooth transition
+between the components of the different repos. You can find more details about the migration in https://github.com/kubeflow/dashboard/issues/118.
+
+### Prerequisites
+
+1. A Kubeflow cluster with 1.10.0 or above
+2. `kubectl` configured for access to the above cluster
+
+### Script
+
+You can use the following script to perform the migration from `v1.10.0` components to the `v2.0.0` ones from this repository. The script will
+perform the following actions:
+1. Remove the existing components of CentralDashboard, Profiles Controller and PodDefaults Webhook
+    * The script will not remove any CR (Custom Resource) or CRD (Custom Resource Definition), to ensure no data loss
+    * Only Kubernetes resources relevant to the Deployments will be removed (Deployment, ServiceAccount, Service etc)
+    * The [`NetworkPolicy`](https://github.com/kubeflow/manifests/blob/v1.10-branch/common/networkpolicies/base/centraldashboard.yaml) from the `kubeflow/manifests` repo, of the CentralDashboard, will be removed
+2. Install the manifests from this repository for the Dashboard, Profiles Controller and PodDefaults webhook
+
+```bash
+./scripts/upgrade_v1_to_v1.sh
+```
+
 ## Community
 
 Kubeflow Dashboard is part of the Kubeflow project, refer to the [Kubeflow Community](https://www.kubeflow.org/docs/about/community/) page for more information.

scripts/upgrade_v1_to_v2.sh

@@ -0,0 +1,97 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
+
+echo "--------------------------------------------------------------------------------------"
+echo "Running the upgrade script to migrate the Kubeflow Dashboard components to V2 release."
+echo "--------------------------------------------------------------------------------------"
+
+PROFILES_LABELS=kustomize.component=profiles
+DASHBOARD_LABELS=app.kubernetes.io/component=centraldashboard
+PODDEFAULT_LABELS=app.kubernetes.io/component=poddefaults
+
+# Helper function for removing all K8s resources of a Kubeflow Component. This will include all resources
+# relevant to the Deployment, but not CRDs. This is done to ensure we don't accidentally delete CRs
+# (like PodDefaults) in user namespace.
+remove-component() {
+  label=$1
+  namespace_resources="deployment service role rolebinding configmap serviceaccount virtualservice authorizationpolicy certificate secret"
+  cluster_resources="clusterrole clusterrolebinding mutatingwebhookconfigurations"
+
+  echo -e "\nWill remove namespaced resources with labels: $label"
+  for resource in $namespace_resources; do
+    echo "Removing all $resource objects..."
+    kubectl delete -n kubeflow -l $label $resource
+    echo "Successfully removed all $resource objects"
+  done
+
+  for resource in $cluster_resources; do
+    echo "Removing all $resource objects..."
+    kubectl delete -l $label $resource
+    echo "Successfully removed all $resource objects"
+  done
+}
+
+
+echo -e "\nRemoving PodDefaults component..."
+remove-component $PODDEFAULT_LABELS
+echo -e "\nSuccessfully removed PodDefaults!"
+
+echo -e "\nRemoving Profiles component..."
+remove-component $PROFILES_LABELS
+echo -e "\nSuccessfully removed Profiles!\n"
+
+echo -e "\nRemoving Centraldashboard component..."
+remove-component $DASHBOARD_LABELS
+echo "Removing NetworkPolicy created by kubeflow/manifests repository..."
+kubectl delete networkpolicy -n kubeflow centraldashboard || echo "No NetworkPolicy from manifests repository found. Continuing..."
+echo -e "\nSuccessfully removed Centraldashboard!"
+
+echo "-----------------------------------------------"
+echo "Installing the updated Dashboard V2 components."
+echo "-----------------------------------------------"
+
+echo -e "\nApplying Dashboard component..."
+echo -e "--------------------------------"
+kustomize build \
+  $SCRIPT_DIR/../components/centraldashboard/manifests/overlays/istio \
+  | kubectl apply -f -
+
+echo "Waiting for Dashboard Deployment to become available..."
+kubectl wait -n kubeflow \
+  --for=condition=Available \
+  deployment \
+  dashboard \
+  --timeout=5m
+echo -e "Successfully applied the Dashboard component!"
+
+echo -e "\nApplying PodDefaults component..."
+echo -e "----------------------------------"
+kustomize build \
+  $SCRIPT_DIR/../components/poddefaults-webhooks/manifests/overlays/cert-manager \
+  | kubectl apply -f -
+
+echo "Waiting for PodDefaults Webhook Deployment to become available..."
+kubectl wait -n kubeflow \
+  --for=condition=Available \
+  deployment \
+  poddefaults-webhook-deployment \
+  --timeout=5m
+echo -e "Successfully applied the PodDefaults component!"
+
+echo -e "\nApplying Profile Controller component..."
+echo -e "-----------------------------------------"
+kustomize build \
+  $SCRIPT_DIR/../components/profile-controller/config/overlays/kubeflow/ \
+  | kubectl apply -f -
+
+echo "Waiting for Profiles Controller Deployment to become available..."
+kubectl wait -n kubeflow \
+  --for=condition=Available \
+  deployment \
+  profiles-deployment \
+  --timeout=5m
+echo -e "Successfully applied the Profile Controller component!"
+
+echo -e "\nSuccessfully applied Dashboard V2 components!\n"