Fix bug 2143 to track errors during installation of Kubernetes and helm

[osm/devops.git] / installers / install_kubeadm_cluster.sh

diff --git a/installers/install_kubeadm_cluster.sh b/installers/install_kubeadm_cluster.sh
index 70f6d9d..6f3aa87 100755 (executable)
--- a/installers/install_kubeadm_cluster.sh
+++ b/installers/install_kubeadm_cluster.sh
@@ -15,10 +15,11 @@
 
 set +eux
 
-#installs kubernetes packages
+K8S_VERSION=1.23.3-00
+
+# installs kubernetes packages
 function install_kube() {
     [ -z "${DEBUG_INSTALL}" ] || DEBUG beginning of function
-    K8S_VERSION=1.23.3-00
     # Kubernetes releases can be found here: https://kubernetes.io/releases/
     # To check other available versions, run the following command
     # curl -s https://packages.cloud.google.com/apt/dists/kubernetes-xenial/main/binary-amd64/Packages | grep Version | awk '{print $2}'
@@ -36,39 +37,63 @@ EOF
     [ -z "${DEBUG_INSTALL}" ] || DEBUG end of function
 }
 
-#initializes kubernetes control plane
+# check and track kube packages installation
+function check_and_track_kube_install() {
+    [ -z "${DEBUG_INSTALL}" ] || DEBUG beginning of function
+    kubelet_version=$(dpkg -s kubelet|grep Version|awk '{print $2}')
+    [ -n "${kubelet_version}" ] || FATAL_TRACK k8scluster "Kubelet was not installed."
+    kubeadm_version=$(dpkg -s kubeadm|grep Version|awk '{print $2}')
+    [ -n "${kubeadm_version}" ] || FATAL_TRACK k8scluster "Kubeadm was not installed."
+    kubectl_version=$(dpkg -s kubectl|grep Version|awk '{print $2}')
+    [ -n "${kubectl_version}" ] || FATAL_TRACK k8scluster "Kubectl was not installed."
+    track k8scluster install_k8s_ok kubelet ${kubelet_version} none none kubeadm ${kubeadm_version} none none kubectl ${kubectl_version} none none
+    [ -z "${DEBUG_INSTALL}" ] || DEBUG end of function
+}
+
+# initializes kubernetes control plane
 function init_kubeadm() {
     [ -z "${DEBUG_INSTALL}" ] || DEBUG beginning of function
     sudo swapoff -a
     sudo sed -i.bak '/.*none.*swap/s/^\(.*\)$/#\1/g' /etc/fstab
+    sudo kubeadm init --dry-run || FATAL_TRACK k8scluster "kubeadm init dry-run failed"
     sudo kubeadm init --config $1
     sleep 5
     [ -z "${DEBUG_INSTALL}" ] || DEBUG end of function
 }
 
+# Initializes kubeconfig file
 function kube_config_dir() {
     [ -z "${DEBUG_INSTALL}" ] || DEBUG beginning of function
     K8S_MANIFEST_DIR="/etc/kubernetes/manifests"
-    [ ! -d $K8S_MANIFEST_DIR ] && FATAL "Cannot Install Kubernetes"
+    [ ! -d $K8S_MANIFEST_DIR ] && FATAL_TRACK k8scluster "Kubernetes folder $K8S_MANIFEST_DIR was not found"
     mkdir -p $HOME/.kube
     sudo cp /etc/kubernetes/admin.conf $HOME/.kube/config
     sudo chown $(id -u):$(id -g) $HOME/.kube/config
     [ -z "${DEBUG_INSTALL}" ] || DEBUG end of function
 }
 
-#deploys flannel as daemonsets
+# test kubernetes installation
+function check_and_track_init_k8s() {
+    [ -z "${DEBUG_INSTALL}" ] || DEBUG beginning of function
+    kubectl get ns || FATAL_TRACK k8scluster "Failed getting namespaces"
+    track k8scluster init_k8s_ok
+    [ -z "${DEBUG_INSTALL}" ] || DEBUG end of function
+}
+
+# deploys flannel as daemonsets
 function deploy_cni_provider() {
     [ -z "${DEBUG_INSTALL}" ] || DEBUG beginning of function
     CNI_DIR="$(mktemp -d -q --tmpdir "flannel.XXXXXX")"
     trap 'rm -rf "${CNI_DIR}"' EXIT
-    wget --retry-on-host-error --retry-on-http-error 404,429,503 --tries=5 https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml -P $CNI_DIR
-    [ ! -f $CNI_DIR/kube-flannel.yml ] && FATAL "Cannot Install Flannel because $CNI_DIR/kube-flannel.yml was not found. Maybe the file https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml is temporarily not accessible"
+    KUBE_FLANNEL_FILE_URL="https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml"
+    wget --retry-on-host-error --retry-on-http-error 404,429,503 --tries=5 "${KUBE_FLANNEL_FILE_URL}" -P $CNI_DIR
+    [ ! -f $CNI_DIR/kube-flannel.yml ] && FATAL_TRACK k8scluster "Cannot Install Flannel because $CNI_DIR/kube-flannel.yml was not found. Maybe the file ${KUBE_FLANNEL_FILE_URL} is temporarily not accessible"
     kubectl apply -f $CNI_DIR
-    [ $? -ne 0 ] && FATAL "Cannot Install Flannel"
+    [ $? -ne 0 ] && FATAL_TRACK k8scluster "Cannot Install Flannel"
     [ -z "${DEBUG_INSTALL}" ] || DEBUG end of function
 }
 
-#taints K8s master node
+# taints K8s master node
 function taint_master_node() {
     [ -z "${DEBUG_INSTALL}" ] || DEBUG beginning of function
     K8S_MASTER=$(kubectl get nodes | awk '$3~/master/'| awk '{print $1}')
@@ -77,6 +102,14 @@ function taint_master_node() {
     [ -z "${DEBUG_INSTALL}" ] || DEBUG end of function
 }
 
+# check and track kube packages installation
+function check_and_track_k8s_ready_before_helm() {
+    [ -z "${DEBUG_INSTALL}" ] || DEBUG beginning of function
+    kubectl get events || FATAL_TRACK k8scluster "Failed getting events"
+    track k8scluster k8s_ready_before_helm
+    [ -z "${DEBUG_INSTALL}" ] || DEBUG end of function
+}
+
 #Install Helm v3
 #Helm releases can be found here: https://github.com/helm/helm/releases
 function install_helm() {
@@ -93,8 +126,9 @@ function install_helm() {
     else
         echo "Helm3 is already installed. Skipping installation..."
     fi
-    helm repo add stable https://charts.helm.sh/stable
-    helm repo update
+    helm version || FATAL_TRACK k8scluster "Could not obtain helm version. Maybe helm client was not installed"
+    helm repo add stable https://charts.helm.sh/stable || FATAL_TRACK k8scluster "Helm repo stable could not be added"
+    helm repo update || FATAL_TRACK k8scluster "Helm repo stable could not be updated"
     [ -z "${DEBUG_INSTALL}" ] || DEBUG end of function
 }
 
@@ -127,7 +161,7 @@ function install_k8s_storageclass() {
             sleep 15
         fi
     done
-    [ -n "$storageclass_ready" ] || FATAL "Storageclass not ready after $storageclass_timeout seconds. Cannot install openebs"
+    [ -n "$storageclass_ready" ] || FATAL_TRACK k8scluster "Storageclass not ready after $storageclass_timeout seconds. Cannot install openebs"
     kubectl patch storageclass openebs-hostpath -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'
     [ -z "${DEBUG_INSTALL}" ] || DEBUG end of function
 }
@@ -226,7 +260,7 @@ function check_for_readiness() {
     if [[ (${failures_in_a_row} -ge ${failures_threshold}) ]]
     then
         echo
-        FATAL "K8S CLUSTER IS BROKEN"
+        FATAL_TRACK k8scluster "K8S CLUSTER IS BROKEN"
     else
         echo
         echo "K8S CLUSTER IS READY"
@@ -284,21 +318,26 @@ echo "HOME=$HOME"
 
 
 install_kube
-track k8scluster install_k8s_ok
+check_and_track_kube_install
+
 init_kubeadm $OSM_DOCKER_WORK_DIR/cluster-config.yaml
 kube_config_dir
-track k8scluster init_k8s_ok
+check_and_track_init_k8s
+
 if [ -n "$INSTALL_K8S_MONITOR" ]; then
     # uninstall OSM MONITORING
     uninstall_k8s_monitoring
     track k8scluster uninstall_k8s_monitoring_ok
 fi
-#remove old namespace
+
 remove_k8s_namespace osm
 deploy_cni_provider
 taint_master_node
+check_and_track_k8s_ready_before_helm
+
 install_helm
 track k8scluster install_helm_ok
+
 install_k8s_storageclass
 track k8scluster k8s_storageclass_ok
 install_helm_metallb