Automated cherry pick of #14684 #14689 #14690 #16719 upstream release 1.1

cluster/gce/config-test.sh

@@ -118,7 +118,7 @@ if [[ "${ENABLE_HORIZONTAL_POD_AUTOSCALER}" == "true" ]]; then
 fi
 
 # Optional: Enable deployment experimental feature, not ready for production use.
-ENABLE_DEPLOYMENTS="${KUBE_ENABLE_DEPLOYMENTS:-false}"
+ENABLE_DEPLOYMENTS="${KUBE_ENABLE_DEPLOYMENTS:-true}"
 if [[ "${ENABLE_DEPLOYMENTS}" == "true" ]]; then
   ENABLE_EXPERIMENTAL_API=true
 fi

cluster/gce/coreos/helper.sh

@@ -148,15 +148,10 @@ function create-master-instance {
 
 # TODO(dawnchen): Check $CONTAINER_RUNTIME to decide which
 # cloud_config yaml file should be passed
-# TODO(zmerlynn): Make $1 required.
-# TODO(zmerlynn): Document required vars (for this and call chain).
-# $1 version
+# $1: template name (required)
 function create-node-instance-template {
-  local suffix=""
-  if [[ -n ${1:-} ]]; then
-    suffix="-${1}"
-  fi
-   create-node-template "${NODE_INSTANCE_PREFIX}-template${suffix}" "${scope_flags}" \
+  local template_name="$1"
+  create-node-template "$template_name" "${scope_flags}" \
     "kube-env=${KUBE_TEMP}/node-kube-env.yaml" \
     "user-data=${KUBE_ROOT}/cluster/gce/coreos/node.yaml"
 }

cluster/gce/debian/helper.sh

@@ -167,15 +167,10 @@ function create-master-instance {
     --disk "name=${MASTER_NAME}-pd,device-name=master-pd,mode=rw,boot=no,auto-delete=no"
 }
 
-# TODO(zmerlynn): Make $1 required.
-# TODO(zmerlynn): Document required vars (for this and call chain).
-# $1 version
+# $1: template name (required)
 function create-node-instance-template {
-  local suffix=""
-  if [[ -n ${1:-} ]]; then
-    suffix="-${1}"
-  fi
-  create-node-template "${NODE_INSTANCE_PREFIX}-template${suffix}" "${scope_flags}" \
+  local template_name="$1"
+  create-node-template "$template_name" "${scope_flags}" \
     "startup-script=${KUBE_ROOT}/cluster/gce/configure-vm.sh" \
     "kube-env=${KUBE_TEMP}/node-kube-env.yaml"
 }

cluster/gce/trusty/helper.sh

@@ -27,15 +27,10 @@
 # create-node-instance-template function to use Ubuntu.
 source "${KUBE_ROOT}/cluster/gce/debian/helper.sh"
 
-# TODO(andyzheng0831): Make $1 required.
-# TODO(andyzheng0831): Document required vars (for this and call chain).
-# $1 version
+# $1: template name (required)
 function create-node-instance-template {
-  local suffix=""
-  if [[ -n ${1:-} ]]; then
-    suffix="-${1}"
-  fi
-  create-node-template "${NODE_INSTANCE_PREFIX}-template${suffix}" "${scope_flags[*]}" \
+  local template_name="$1"
+  create-node-template "$template_name" "${scope_flags[*]}" \
 		"kube-env=${KUBE_TEMP}/node-kube-env.yaml" \
     "user-data=${KUBE_ROOT}/cluster/gce/trusty/node.yaml"
 }

cluster/gce/upgrade.sh

@@ -173,7 +173,7 @@ function upgrade-nodes() {
 #   KUBELET_KEY_BASE64
 function prepare-node-upgrade() {
   echo "== Preparing node upgrade (to ${KUBE_VERSION}). ==" >&2
-  SANITIZED_VERSION=$(echo ${KUBE_VERSION} | sed s/"\."/-/g)
+  SANITIZED_VERSION=$(echo ${KUBE_VERSION} | sed 's/[\.\+]/-/g')
 
   detect-minion-names
 
@@ -200,9 +200,10 @@ function prepare-node-upgrade() {
 
   # TODO(zmerlynn): Get configure-vm script from ${version}. (Must plumb this
   #                 through all create-node-instance-template implementations).
-  create-node-instance-template ${SANITIZED_VERSION}
+  local template_name=$(get-template-name-from-version ${SANITIZED_VERSION})
+  create-node-instance-template "${template_name}"
   # The following is echo'd so that callers can get the template name.
-  echo "${NODE_INSTANCE_PREFIX}-template-${SANITIZED_VERSION}"
+  echo $template_name
   echo "== Finished preparing node upgrade (to ${KUBE_VERSION}). ==" >&2
 }
 
@@ -220,12 +221,13 @@ function do-node-upgrade() {
   if [[ "${exists}" != "0" ]]; then
     subgroup="alpha compute"
   fi
+  local template_name=$(get-template-name-from-version ${SANITIZED_VERSION})
   gcloud ${subgroup} rolling-updates \
       --project="${PROJECT}" \
       --zone="${ZONE}" \
       start \
       --group="${NODE_INSTANCE_PREFIX}-group" \
-      --template="${NODE_INSTANCE_PREFIX}-template-${SANITIZED_VERSION}" \
+      --template="${template_name}" \
       --instance-startup-timeout=300s \
       --max-num-concurrent-instances=1 \
       --max-num-failed-instances=0 \

cluster/gce/util.sh

@@ -345,21 +345,28 @@ function create-firewall-rule {
   done
 }
 
+# $1: version (required)
+function get-template-name-from-version {
+  # trim template name to pass gce name validation
+  echo "${NODE_INSTANCE_PREFIX}-template-${1}" | cut -c 1-63 | sed 's/[\.\+]/-/g;s/-*$//g'
+}
+
 # Robustly try to create an instance template.
 # $1: The name of the instance template.
 # $2: The scopes flag.
 # $3: The minion start script metadata from file.
 # $4: The kube-env metadata.
 function create-node-template {
   detect-project
+  local template_name="$1"
 
   # First, ensure the template doesn't exist.
   # TODO(zmerlynn): To make this really robust, we need to parse the output and
   #                 add retries. Just relying on a non-zero exit code doesn't
   #                 distinguish an ephemeral failed call from a "not-exists".
-  if gcloud compute instance-templates describe "$1" --project "${PROJECT}" &>/dev/null; then
+  if gcloud compute instance-templates describe "$template_name" --project "${PROJECT}" &>/dev/null; then
     echo "Instance template ${1} already exists; deleting." >&2
-    if ! gcloud compute instance-templates delete "$1" --project "${PROJECT}" &>/dev/null; then
+    if ! gcloud compute instance-templates delete "$template_name" --project "${PROJECT}" &>/dev/null; then
       echo -e "${color_yellow}Failed to delete existing instance template${color_norm}" >&2
       exit 2
     fi
@@ -372,7 +379,7 @@ function create-node-template {
   fi
   while true; do
     echo "Attempt ${attempt} to create ${1}" >&2
-    if ! gcloud compute instance-templates create "$1" \
+    if ! gcloud compute instance-templates create "$template_name" \
       --project "${PROJECT}" \
       --machine-type "${MINION_SIZE}" \
       --boot-disk-type "${MINION_DISK_TYPE}" \
@@ -386,10 +393,10 @@ function create-node-template {
       --can-ip-forward \
       --metadata-from-file "$3","$4" >&2; then
         if (( attempt > 5 )); then
-          echo -e "${color_red}Failed to create instance template $1 ${color_norm}" >&2
+          echo -e "${color_red}Failed to create instance template $template_name ${color_norm}" >&2
           exit 2
         fi
-        echo -e "${color_yellow}Attempt ${attempt} failed to create instance template $1. Retrying.${color_norm}" >&2
+        echo -e "${color_yellow}Attempt ${attempt} failed to create instance template $template_name. Retrying.${color_norm}" >&2
         attempt=$(($attempt+1))
     else
         break
@@ -663,15 +670,18 @@ function kube-up {
   fi
 
   write-node-env
-  create-node-instance-template
+
+  local template_name="${NODE_INSTANCE_PREFIX}-template"
+
+  create-node-instance-template $template_name
 
   gcloud compute instance-groups managed \
       create "${NODE_INSTANCE_PREFIX}-group" \
       --project "${PROJECT}" \
       --zone "${ZONE}" \
       --base-instance-name "${NODE_INSTANCE_PREFIX}" \
       --size "${NUM_MINIONS}" \
-      --template "${NODE_INSTANCE_PREFIX}-template" || true;
+      --template "$template_name" || true;
   gcloud compute instance-groups managed wait-until-stable \
       "${NODE_INSTANCE_PREFIX}-group" \
 			--zone "${ZONE}" \
@@ -1034,31 +1044,33 @@ function prepare-push() {
 
     # Ugly hack: Since it is not possible to delete instance-template that is currently
     # being used, create a temp one, then delete the old one and recreate it once again.
-    create-node-instance-template "tmp"
+    local tmp_template_name="${NODE_INSTANCE_PREFIX}-template-tmp"
+    create-node-instance-template $tmp_template_name
 
+    local template_name="${NODE_INSTANCE_PREFIX}-template"
     gcloud compute instance-groups managed \
       set-instance-template "${NODE_INSTANCE_PREFIX}-group" \
-      --template "${NODE_INSTANCE_PREFIX}-template-tmp" \
+      --template "$tmp_template_name" \
       --zone "${ZONE}" \
       --project "${PROJECT}" || true;
 
     gcloud compute instance-templates delete \
       --project "${PROJECT}" \
       --quiet \
-      "${NODE_INSTANCE_PREFIX}-template" || true
+      "$template_name" || true
 
-    create-node-instance-template
+    create-node-instance-template "$template_name"
 
     gcloud compute instance-groups managed \
       set-instance-template "${NODE_INSTANCE_PREFIX}-group" \
-      --template "${NODE_INSTANCE_PREFIX}-template" \
+      --template "$template_name" \
       --zone "${ZONE}" \
       --project "${PROJECT}" || true;
 
     gcloud compute instance-templates delete \
       --project "${PROJECT}" \
       --quiet \
-      "${NODE_INSTANCE_PREFIX}-template-tmp" || true
+      "$tmp_template_name" || true
   fi
 }
 

test/e2e/cluster_upgrade.go

@@ -522,8 +522,8 @@ func migTemplate() (string, error) {
 		// An `instance-groups managed describe` call outputs what we want to stdout.
 		output, _, err := retryCmd("gcloud", "compute", "instance-groups", "managed",
 			fmt.Sprintf("--project=%s", testContext.CloudConfig.ProjectID),
-			fmt.Sprintf("--zone=%s", testContext.CloudConfig.Zone),
 			"describe",
+			fmt.Sprintf("--zone=%s", testContext.CloudConfig.Zone),
 			testContext.CloudConfig.NodeInstanceGroup)
 		if err != nil {
 			errLast = fmt.Errorf("gcloud compute instance-groups managed describe call failed with err: %v", err)