Merge pull request #3034 from a-robinson/pd-up

Store all of the master's persistent data on a separate "data" PD when running on GCE

cluster/gce/templates/format-and-mount-pd.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+
+# Copyright 2014 Google Inc. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# Formats and mounts a persistent disk to store the persistent data on the
+# master -- etcd's data and the security certs/keys.
+
+device_info=$(ls -l /dev/disk/by-id/google-master-pd)
+relative_path=${device_info##* }
+device_path="/dev/disk/by-id/${relative_path}"
+
+# Format and mount the disk to the directory used by etcd.
+mkdir -p /mnt/master-pd
+/usr/share/google/safe_format_and_mount -m "mkfs.ext4 -F" "${device_path}" /mnt/master-pd
+mkdir -m 700 -p /mnt/master-pd/var/etcd
+mkdir -p /mnt/master-pd/srv/kubernetes
+ln -s /mnt/master-pd/var/etcd /var/etcd
+ln -s /mnt/master-pd/srv/kubernetes /srv/kubernetes
+
+# This is a bit of a hack to get around the fact that salt has to run after the
+# PD and mounted directory are already set up. We can't give ownership of the
+# directory to etcd until the etcd user and group exist, but they don't exist
+# until salt runs if we don't create them here. We could alternatively make the
+# permissions on the directory more permissive, but this seems less bad.
+useradd -s /sbin/nologin -d /var/etcd etcd
+chown etcd /mnt/master-pd/var/etcd
+chgrp etcd /mnt/master-pd/var/etcd

cluster/gce/util.sh

@@ -393,6 +393,7 @@ function kube-up {
     echo "readonly ENABLE_NODE_LOGGING='${ENABLE_NODE_LOGGING:-false}'"
     echo "readonly LOGGING_DESTINATION='${LOGGING_DESTINATION:-}'"
     grep -v "^#" "${KUBE_ROOT}/cluster/gce/templates/common.sh"
+    grep -v "^#" "${KUBE_ROOT}/cluster/gce/templates/format-and-mount-pd.sh"
     grep -v "^#" "${KUBE_ROOT}/cluster/gce/templates/create-dynamic-salt-files.sh"
     grep -v "^#" "${KUBE_ROOT}/cluster/gce/templates/download-release.sh"
     grep -v "^#" "${KUBE_ROOT}/cluster/gce/templates/salt-master.sh"
@@ -407,6 +408,13 @@ function kube-up {
     fi
   fi
 
+  # We have to make sure the disk is created before creating the master VM, so
+  # run this in the foreground.
+  gcloud compute disks create "${MASTER_NAME}-pd" \
+    --project "${PROJECT}" \
+    --zone "${ZONE}" \
+    --size "10GB"
+
   gcloud compute instances create "${MASTER_NAME}" \
     --project "${PROJECT}" \
     --zone "${ZONE}" \
@@ -416,7 +424,8 @@ function kube-up {
     --tags "${MASTER_TAG}" \
     --network "${NETWORK}" \
     --scopes "storage-ro" "compute-rw" \
-    --metadata-from-file "startup-script=${KUBE_TEMP}/master-start.sh" &
+    --metadata-from-file "startup-script=${KUBE_TEMP}/master-start.sh" \
+    --disk name="${MASTER_NAME}-pd" device-name=master-pd mode=rw boot=no auto-delete=no &
 
   # Create a single firewall rule for all minions.
   create-firewall-rule "${MINION_TAG}-all" "${CLUSTER_IP_RANGE}" "${MINION_TAG}" &