ci: improve the ci for webhook installation

Signed-off-by: Vicente Cheng <vicente.cheng@suse.com>

.github/workflows/basic-ci.yaml

@@ -59,9 +59,9 @@ jobs:
       - name: "Deploy NDM"
         run: |
           pushd ndm-new-vagrant-k3s
-          cp ../ci/scripts/deploy_ndm.sh ./deploy_ndm.sh
+          cp ../ci/scripts/deploy_ndm_current.sh ./deploy_ndm_current.sh
           cp ../ci/charts/ndm-override.yaml ./ndm-override.yaml
-          ./deploy_ndm.sh
+          ./deploy_ndm_current.sh
           popd
       - name: "Add disk"
         run: |
@@ -123,8 +123,8 @@ jobs:
           echo "VM_DEPLOYED=true" >> "$GITHUB_ENV"
           yq e -i ".longhorn_version = \"1.7.1\"" settings.yaml
           ./scripts/deploy_longhorn.sh
-          cp ../ci/scripts/deploy_ndm.sh ./deploy_ndm.sh
-          ./deploy_ndm.sh
+          cp ../ci/scripts/deploy_ndm_chart.sh ./deploy_ndm_chart.sh
+          ./deploy_ndm_chart.sh
           popd
       - name: "Add disk"
         run: |

ci/charts/ndm-override.yaml

@@ -4,6 +4,12 @@ image:
   # Overrides the image tag whose default is the chart appVersion.
   tag: ""
 
+webhook:
+  repository: rancher/harvester-node-disk-manager
+  pullPolicy: IfNotPresent
+  # Overrides the image tag whose default is the chart appVersion.
+  tag: ""
+
 autoProvisionFilter: [/dev/sd*]
 debug: true
 

ci/scripts/deploy_ndm_current.sh

@@ -0,0 +1,91 @@
+#!/bin/bash -e
+
+TOP_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )/" &> /dev/null && pwd )"
+
+ensure_command() {
+  local cmd=$1
+  if ! which $cmd &> /dev/null; then
+    echo 1
+    return
+  fi
+  echo 0
+}
+
+wait_ndm_ready() {
+  while [ true ]; do
+    running_num=$(kubectl get ds harvester-node-disk-manager -n harvester-system -o 'jsonpath={.status.numberReady}')
+    if [[ $running_num -eq ${cluster_nodes} ]]; then
+      echo "harvester-node-disk-manager pods are ready!"
+      break
+    fi
+    echo "harvester-node-disk-manager pods are not ready, sleep 10 seconds."
+    sleep 10
+  done
+}
+
+ensure_longhorn_ready() {
+  # ensure longhorn-manager first
+  while [ true ]; do
+    running_num=$(kubectl get ds longhorn-manager -n longhorn-system -o 'jsonpath={.status.numberReady}')
+    if [[ $running_num -eq ${cluster_nodes} ]]; then
+      echo "longhorn-manager pods are ready!"
+      break
+    fi
+    echo "check longhorn-manager failure, please deploy longhorn first."
+    exit 1
+  done
+
+  # ensure instance-manager-e ready
+  while [ true ]; do
+    running_num=$(kubectl get pods -n longhorn-system |grep ^instance-manager |grep Running |awk '{print $3}' |wc -l)
+    if [[ $running_num -eq ${cluster_nodes} ]]; then
+      echo "instance-manager pods are ready!"
+      break
+    fi
+    echo "check instance-manager failure, please deploy longhorn first."
+    exit 1
+  done
+}
+
+if [ ! -f $TOP_DIR/kubeconfig ]; then
+  echo "kubeconfig does not exist. Please create cluster first."
+  echo "Maybe try new_cluster.sh"
+  exit 1
+fi
+echo $TOP_DIR/kubeconfig
+export KUBECONFIG=$TOP_DIR/kubeconfig
+
+if [[ $(ensure_command helm) -eq 1 ]]; then
+  echo "no helm, try to curl..."
+  curl -O https://get.helm.sh/helm-v3.9.4-linux-amd64.tar.gz
+  tar -zxvf helm-v3.9.4-linux-amd64.tar.gz
+  HELM=$TOP_DIR/linux-amd64/helm
+  $HELM version
+else
+  echo "Get helm, version info as below"
+  HELM=$(which helm)
+  $HELM version
+fi
+
+cluster_nodes=$(yq -e e '.cluster_size' $TOP_DIR/settings.yaml)
+echo "cluster nodes: $cluster_nodes"
+ensure_longhorn_ready
+
+pushd $TOP_DIR
+cat >> ndm-override.yaml.default << 'EOF'
+autoProvisionFilter: [/dev/sd*]
+EOF
+
+if [ ! -f ndm-override.yaml ]; then
+  mv ndm-override.yaml.default ndm-override.yaml
+fi
+
+cp -r ../deploy/charts/harvester-node-disk-manager harvester-node-disk-manager
+
+target_img=$(yq -e .image.repository ndm-override.yaml)
+echo "install target image: ${target_img}"
+$HELM install -f $TOP_DIR/ndm-override.yaml harvester-node-disk-manager ./harvester-node-disk-manager --create-namespace -n harvester-system
+
+wait_ndm_ready
+echo "harvester-node-disk-manager is ready"
+popd

ci/scripts/get-debug-info.sh

@@ -4,7 +4,7 @@ TARGETNODE="node1"
 
 export KUBECONFIG=kubeconfig
 
-NDMPOD=$(kubectl get pods -n harvester-system --field-selector spec.nodeName=$TARGETNODE |grep ^harvester-node-disk-manager |awk '{print $1}')
+NDMPOD=$(kubectl get pods -n harvester-system --field-selector spec.nodeName=$TARGETNODE | grep ^harvester-node-disk-manager |grep -v webhook |awk '{print $1}')
 
 # filter out the redundant Skip log
 kubectl logs $NDMPOD -n harvester-system |grep -v Skip

ci/scripts/patch-ttl-repo.sh

@@ -2,6 +2,9 @@
 
 COMMIT=$(git rev-parse --short HEAD)
 IMAGE=ttl.sh/node-disk-manager-${COMMIT}
+IMAGE_WEBHOOK=ttl.sh/node-disk-manager-webhook-${COMMIT}
 
 yq e -i ".image.repository = \"${IMAGE}\"" ci/charts/ndm-override.yaml
-yq e -i ".image.tag = \"1h\"" ci/charts/ndm-override.yaml 
+yq e -i ".image.tag = \"1h\"" ci/charts/ndm-override.yaml 
+yq e -i ".webhook.image.repository = \"${IMAGE_WEBHOOK}\"" ci/charts/ndm-override.yaml
+yq e -i ".webhook.image.tag = \"1h\"" ci/charts/ndm-override.yaml 

ci/scripts/upgrade_ndm.sh

@@ -62,7 +62,7 @@ sleep 30 # wait 30 seconds for ndm respawn pods
 
 wait_ndm_ready
 # check image
-pod_name=$(kubectl get pods -n harvester-system |grep Running |grep ^harvester-node-disk-manager|head -n1 |awk '{print $1}')
+pod_name=$(kubectl get pods -n harvester-system |grep Running |grep -v webhook |grep ^harvester-node-disk-manager|head -n1 |awk '{print $1}')
 container_img=$(kubectl get pods ${pod_name} -n harvester-system -o yaml |yq -e .spec.containers[0].image |tr ":" \n)
 yaml_img=$(yq -e .image.repository ndm-override.yaml)
 if grep -q ${yaml_img} <<< ${container_img}; then