devstack_custom/stack.sh

#!/usr/bin/env bash

# **stack.sh** is an opinionated openstack developer installation.

# This script installs and configures *nova*, *glance*, *dashboard* and *keystone*

# This script allows you to specify configuration options of what git
# repositories to use, enabled services, network configuration and various
# passwords.  If you are crafty you can run the script on multiple nodes using
# shared settings for common resources (mysql, rabbitmq) and build a multi-node
# developer install.

# To keep this script simple we assume you are running on an **Ubuntu 11.04
# Natty** machine.  It should work in a VM or physical server.  Additionally we
# put the list of *apt* and *pip* dependencies and other configuration files in
# this repo.  So start by grabbing this script and the dependencies.

# Learn more and get the most recent version at http://devstack.org

# Sanity Check
# ============

# Warn users who aren't on natty, but allow them to override check and attempt
# installation with ``FORCE=yes ./stack``
if ! grep -q natty /etc/lsb-release; then
    echo "WARNING: this script has only been tested on natty"
    if [[ "$FORCE" != "yes" ]]; then
        echo "If you wish to run this script anyway run with FORCE=yes"
        exit 1
    fi
fi

# Keep track of the current devstack directory.
TOP_DIR=$(cd $(dirname "$0") && pwd)

# stack.sh keeps the list of **apt** and **pip** dependencies in external
# files, along with config templates and other useful files.  You can find these
# in the ``files`` directory (next to this script).  We will reference this
# directory using the ``FILES`` variable in this script.
FILES=$TOP_DIR/files
if [ ! -d $FILES ]; then
    echo "ERROR: missing devstack/files - did you grab more than just stack.sh?"
    exit 1
fi



# Settings
# ========

# This script is customizable through setting environment variables.  If you
# want to override a setting you can either::
#
#     export MYSQL_PASSWORD=anothersecret
#     ./stack.sh
#
# You can also pass options on a single line ``MYSQL_PASSWORD=simple ./stack.sh``
#
# Additionally, you can put any local variables into a ``localrc`` file, like::
#
#     MYSQL_PASSWORD=anothersecret
#     MYSQL_USER=hellaroot
#
# We try to have sensible defaults, so you should be able to run ``./stack.sh``
# in most cases.
#
# We our settings from ``stackrc``.  This file is distributed with devstack and
# contains locations for what repositories to use.  If you want to use other
# repositories and branches, you can add your own settings with another file
# called ``localrc``
#
# If ``localrc`` exists, then ``stackrc`` will load those settings.  This is
# useful for changing a branch or repostiory to test other versions.  Also you
# can store your other settings like **MYSQL_PASSWORD** or **ADMIN_PASSWORD** instead
# of letting devstack generate random ones for you.
source ./stackrc

# Destination path for installation ``DEST``
DEST=${DEST:-/opt/stack}

# OpenStack is designed to be run as a regular user (Dashboard will fail to run
# as root, since apache refused to startup serve content from root user).  If
# stack.sh is run as root, it automatically creates a stack user with
# sudo privileges and runs as that user.

if [[ $EUID -eq 0 ]]; then
    echo "You are running this script as root."
    echo "In 10 seconds, we will create a user 'stack' and run as that user"
    sleep 10

    # since this script runs as a normal user, we need to give that user
    # ability to run sudo
    apt_get update
    apt_get install sudo

    if ! getent passwd stack >/dev/null; then
        echo "Creating a user called stack"
        useradd -U -G sudo -s /bin/bash -d $DEST -m stack
    fi

    echo "Giving stack user passwordless sudo priviledges"
    # natty uec images sudoers does not have a '#includedir'. add one.
    grep -q "^#includedir.*/etc/sudoers.d" /etc/sudoers ||
        echo "#includedir /etc/sudoers.d" >> /etc/sudoers
    ( umask 226 && echo "stack ALL=(ALL) NOPASSWD:ALL" \
        > /etc/sudoers.d/50_stack_sh )

    echo "Copying files to stack user"
    STACK_DIR="$DEST/${PWD##*/}"
    cp -r -f "$PWD" "$STACK_DIR"
    chown -R stack "$STACK_DIR"
    if [[ "$SHELL_AFTER_RUN" != "no" ]]; then
        exec su -c "set -e; cd $STACK_DIR; bash stack.sh; bash" stack
    else
        exec su -c "set -e; cd $STACK_DIR; bash stack.sh" stack
    fi
    exit 1
fi

# Set the destination directories for openstack projects
NOVA_DIR=$DEST/nova
DASH_DIR=$DEST/dash
GLANCE_DIR=$DEST/glance
KEYSTONE_DIR=$DEST/keystone
NOVACLIENT_DIR=$DEST/python-novaclient
OPENSTACKX_DIR=$DEST/openstackx
NOVNC_DIR=$DEST/noVNC

# Specify which services to launch.  These generally correspond to screen tabs
ENABLED_SERVICES=${ENABLED_SERVICES:-g-api,g-reg,key,n-api,n-cpu,n-net,n-sch,n-vnc,dash,mysql,rabbit}

# Nova hypervisor configuration.  We default to **kvm** but will drop back to
# **qemu** if we are unable to load the kvm module.  Stack.sh can also install
# an **LXC** based system.
LIBVIRT_TYPE=${LIBVIRT_TYPE:-kvm}

# nova supports pluggable schedulers.  ``SimpleScheduler`` should work in most
# cases unless you are working on multi-zone mode.
SCHEDULER=${SCHEDULER:-nova.scheduler.simple.SimpleScheduler}

# Use the first IP unless an explicit is set by ``HOST_IP`` environment variable
if [ ! -n "$HOST_IP" ]; then
    HOST_IP=`LC_ALL=C /sbin/ifconfig  | grep -m 1 'inet addr:'| cut -d: -f2 | awk '{print $1}'`
fi

# apt-get wrapper to just get arguments set correctly
function apt_get() {
    local sudo="sudo"
    [ "$(id -u)" = "0" ] && sudo=""
    $sudo DEBIAN_FRONTEND=noninteractive apt-get \
        --option "Dpkg::Options::=--force-confold" --assume-yes "$@"
}

# Generic helper to configure passwords
function read_password {
    set +o xtrace
    var=$1; msg=$2
    pw=${!var}

    localrc=$TOP_DIR/localrc

    # If the password is not defined yet, proceed to prompt user for a password.
    if [ ! $pw ]; then
        # If there is no localrc file, create one
        if [ ! -e $localrc ]; then
            touch $localrc
        fi

        # Presumably if we got this far it can only be that our localrc is missing
        # the required password.  Prompt user for a password and write to localrc.
        echo ''
        echo '################################################################################'
        echo $msg
        echo '################################################################################'
        echo "This value will be written to your localrc file so you don't have to enter it again."
        echo "It is probably best to avoid spaces and weird characters."
        echo "If you leave this blank, a random default value will be used."
        echo "Enter a password now:"
        read $var
        pw=${!var}
        if [ ! $pw ]; then
            pw=`openssl rand -hex 10`
        fi
        eval "$var=$pw"
        echo "$var=$pw" >> $localrc
    fi
    set -o xtrace
}


# Nova Network Configuration
# --------------------------

# FIXME: more documentation about why these are important flags.  Also
# we should make sure we use the same variable names as the flag names.

PUBLIC_INTERFACE=${PUBLIC_INTERFACE:-eth0}
FIXED_RANGE=${FIXED_RANGE:-10.0.0.0/24}
FIXED_NETWORK_SIZE=${FIXED_NETWORK_SIZE:-256}
FLOATING_RANGE=${FLOATING_RANGE:-172.24.4.1/28}
NET_MAN=${NET_MAN:-FlatDHCPManager}
EC2_DMZ_HOST=${EC2_DMZ_HOST:-$HOST_IP}
FLAT_NETWORK_BRIDGE=${FLAT_NETWORK_BRIDGE:-br100}
VLAN_INTERFACE=${VLAN_INTERFACE:-$PUBLIC_INTERFACE}

# Multi-host is a mode where each compute node runs its own network node.  This
# allows network operations and routing for a VM to occur on the server that is
# running the VM - removing a SPOF and bandwidth bottleneck.
MULTI_HOST=${MULTI_HOST:-0}

# If you are using FlatDHCP on multiple hosts, set the ``FLAT_INTERFACE``
# variable but make sure that the interface doesn't already have an
# ip or you risk breaking things.
#
# **DHCP Warning**:  If your flat interface device uses DHCP, there will be a
# hiccup while the network is moved from the flat interface to the flat network
# bridge.  This will happen when you launch your first instance.  Upon launch
# you will lose all connectivity to the node, and the vm launch will probably
# fail.
#
# If you are running on a single node and don't need to access the VMs from
# devices other than that node, you can set the flat interface to the same
# value as ``FLAT_NETWORK_BRIDGE``.  This will stop the network hiccup from
# occuring.
FLAT_INTERFACE=${FLAT_INTERFACE:-eth0}

## FIXME(ja): should/can we check that FLAT_INTERFACE is sane?


# MySQL & RabbitMQ
# ----------------

# We configure Nova, Dashboard, Glance and Keystone to use MySQL as their
# database server.  While they share a single server, each has their own
# database and tables.

# By default this script will install and configure MySQL.  If you want to
# use an existing server, you can pass in the user/password/host parameters.
# You will need to send the same ``MYSQL_PASSWORD`` to every host if you are doing
# a multi-node devstack installation.
MYSQL_HOST=${MYSQL_HOST:-localhost}
MYSQL_USER=${MYSQL_USER:-root}
read_password MYSQL_PASSWORD "ENTER A PASSWORD TO USE FOR MYSQL."

# don't specify /db in this string, so we can use it for multiple services
BASE_SQL_CONN=${BASE_SQL_CONN:-mysql://$MYSQL_USER:$MYSQL_PASSWORD@$MYSQL_HOST}

# Rabbit connection info
RABBIT_HOST=${RABBIT_HOST:-localhost}
read_password RABBIT_PASSWORD "ENTER A PASSWORD TO USE FOR RABBIT."

# Glance connection info.  Note the port must be specified.
GLANCE_HOSTPORT=${GLANCE_HOSTPORT:-$HOST_IP:9292}


# Keystone
# --------

# Service Token - Openstack components need to have an admin token
# to validate user tokens.
read_password SERVICE_TOKEN "ENTER A SERVICE_TOKEN TO USE FOR THE SERVICE ADMIN TOKEN."
# Dash currently truncates usernames and passwords at 20 characters
read_password ADMIN_PASSWORD "ENTER A PASSWORD TO USE FOR DASH AND KEYSTONE (20 CHARS OR LESS)."

LOGFILE=${LOGFILE:-"$PWD/stack.sh.$$.log"}
(
# So that errors don't compound we exit on any errors so you see only the
# first error that occured.
trap failed ERR
failed() {
    local r=$?
    set +o xtrace
    [ -n "$LOGFILE" ] && echo "${0##*/} failed: full log in $LOGFILE"
    exit $r
}

# Print the commands being run so that we can see the command that triggers
# an error.  It is also useful for following along as the install occurs.
set -o xtrace

# create the destination directory and ensure it is writable by the user
sudo mkdir -p $DEST
if [ ! -w $DEST ]; then
    sudo chown `whoami` $DEST
fi

# Install Packages
# ================
#
# Openstack uses a fair number of other projects.


# install apt requirements
apt_get update
apt_get install `cat $FILES/apts/* | cut -d\# -f1 | grep -Ev "mysql-server|rabbitmq-server"`

# install python requirements
sudo PIP_DOWNLOAD_CACHE=/var/cache/pip pip install `cat $FILES/pips/*`

# git clone only if directory doesn't exist already.  Since ``DEST`` might not
# be owned by the installation user, we create the directory and change the
# ownership to the proper user.
function git_clone {
    # if there is an existing checkout, move it out of the way
    if [[ "$RECLONE" == "yes" ]]; then
        # FIXME(ja): if we were smarter we could speed up RECLONE by
        # using the old git repo as the basis of our new clone...
        if [ -d $2 ]; then
            mv $2 /tmp/stack.`date +%s`
        fi
    fi

    if [ ! -d $2 ]; then
        git clone $1 $2
        cd $2
        # This checkout syntax works for both branches and tags
        git checkout $3
    fi
}

# compute service
git_clone $NOVA_REPO $NOVA_DIR $NOVA_BRANCH
# image catalog service
git_clone $GLANCE_REPO $GLANCE_DIR $GLANCE_BRANCH
# unified auth system (manages accounts/tokens)
git_clone $KEYSTONE_REPO $KEYSTONE_DIR $KEYSTONE_BRANCH
# a websockets/html5 or flash powered VNC console for vm instances
git_clone $NOVNC_REPO $NOVNC_DIR $NOVNC_BRANCH
# django powered web control panel for openstack
git_clone $DASH_REPO $DASH_DIR $DASH_BRANCH $DASH_TAG
# python client library to nova that dashboard (and others) use
git_clone $NOVACLIENT_REPO $NOVACLIENT_DIR $NOVACLIENT_BRANCH
# openstackx is a collection of extensions to openstack.compute & nova
# that is *deprecated*.  The code is being moved into python-novaclient & nova.
git_clone $OPENSTACKX_REPO $OPENSTACKX_DIR $OPENSTACKX_BRANCH

# Initialization
# ==============


# setup our checkouts so they are installed into python path
# allowing ``import nova`` or ``import glance.client``
cd $KEYSTONE_DIR; sudo python setup.py develop
cd $GLANCE_DIR; sudo python setup.py develop
cd $NOVACLIENT_DIR; sudo python setup.py develop
cd $NOVA_DIR; sudo python setup.py develop
cd $OPENSTACKX_DIR; sudo python setup.py develop
cd $DASH_DIR/django-openstack; sudo python setup.py develop
cd $DASH_DIR/openstack-dashboard; sudo python setup.py develop

# Add a useful screenrc.  This isn't required to run openstack but is we do
# it since we are going to run the services in screen for simple
cp $FILES/screenrc ~/.screenrc

## TODO: update current user to allow sudo for all commands in files/sudo/*

# Rabbit
# ---------

if [[ "$ENABLED_SERVICES" =~ "rabbit" ]]; then
    # Install and start rabbitmq-server
    # the temp file is necessary due to LP: #878600
    tfile=$(mktemp)
    apt_get install rabbitmq-server > "$tfile" 2>&1
    cat "$tfile"
    rm -f "$tfile"
    # change the rabbit password since the default is "guest"
    sudo rabbitmqctl change_password guest $RABBIT_PASSWORD
fi

# Mysql
# ---------

if [[ "$ENABLED_SERVICES" =~ "mysql" ]]; then

    # Seed configuration with mysql password so that apt-get install doesn't
    # prompt us for a password upon install.
    cat <<MYSQL_PRESEED | sudo debconf-set-selections
mysql-server-5.1 mysql-server/root_password password $MYSQL_PASSWORD
mysql-server-5.1 mysql-server/root_password_again password $MYSQL_PASSWORD
mysql-server-5.1 mysql-server/start_on_boot boolean true
MYSQL_PRESEED

    # while ``.my.cnf`` is not needed for openstack to function, it is useful
    # as it allows you to access the mysql databases via ``mysql nova`` instead
    # of having to specify the username/password each time.
    if [[ ! -e $HOME/.my.cnf ]]; then
        cat <<EOF >$HOME/.my.cnf
[client]
user=$MYSQL_USER
password=$MYSQL_PASSWORD
host=$MYSQL_HOST
EOF
        chmod 0600 $HOME/.my.cnf
    fi

    # Install and start mysql-server
    apt_get install mysql-server
    # Update the DB to give user ‘$MYSQL_USER’@’%’ full control of the all databases:
    sudo mysql -uroot -p$MYSQL_PASSWORD -e "GRANT ALL PRIVILEGES ON *.* TO '$MYSQL_USER'@'%' identified by '$MYSQL_PASSWORD';"

    # Edit /etc/mysql/my.cnf to change ‘bind-address’ from localhost (127.0.0.1) to any (0.0.0.0) and restart the mysql service:
    sudo sed -i 's/127.0.0.1/0.0.0.0/g' /etc/mysql/my.cnf
    sudo service mysql restart
fi


# Dashboard
# ---------

# Setup the django dashboard application to serve via apache/wsgi

if [[ "$ENABLED_SERVICES" =~ "dash" ]]; then

    # Dash currently imports quantum even if you aren't using it.  Instead
    # of installing quantum we can create a simple module that will pass the
    # initial imports
    mkdir -p  $DASH_DIR/openstack-dashboard/quantum || true
    touch $DASH_DIR/openstack-dashboard/quantum/__init__.py
    touch $DASH_DIR/openstack-dashboard/quantum/client.py


    # ``local_settings.py`` is used to override dashboard default settings.
    cp $FILES/dash_settings.py $DASH_DIR/openstack-dashboard/local/local_settings.py

    # Initialize the dashboard database (it stores sessions and notices shown to
    # users).  The user system is external (keystone).
    cd $DASH_DIR/openstack-dashboard
    dashboard/manage.py syncdb

    # create an empty directory that apache uses as docroot
    sudo mkdir -p $DASH_DIR/.blackhole

    ## Configure apache's 000-default to run dashboard
    sudo cp $FILES/000-default.template /etc/apache2/sites-enabled/000-default
    sudo sed -e "s,%USER%,$USER,g" -i /etc/apache2/sites-enabled/000-default
    sudo sed -e "s,%DASH_DIR%,$DASH_DIR,g" -i /etc/apache2/sites-enabled/000-default
fi


# Glance
# ------

if [[ "$ENABLED_SERVICES" =~ "g-reg" ]]; then
    GLANCE_IMAGE_DIR=$DEST/glance/images
    # Delete existing images
    rm -rf $GLANCE_IMAGE_DIR

    # Use local glance directories
    mkdir -p $GLANCE_IMAGE_DIR

    # (re)create glance database
    mysql -u$MYSQL_USER -p$MYSQL_PASSWORD -e 'DROP DATABASE IF EXISTS glance;'
    mysql -u$MYSQL_USER -p$MYSQL_PASSWORD -e 'CREATE DATABASE glance;'

    # Copy over our glance configurations and update them
    GLANCE_CONF=$GLANCE_DIR/etc/glance-registry.conf
    cp $FILES/glance-registry.conf $GLANCE_CONF
    sudo sed -e "s,%SQL_CONN%,$BASE_SQL_CONN/glance,g" -i $GLANCE_CONF
    sudo sed -e "s,%SERVICE_TOKEN%,$SERVICE_TOKEN,g" -i $GLANCE_CONF
    sudo sed -e "s,%DEST%,$DEST,g" -i $GLANCE_CONF

    GLANCE_API_CONF=$GLANCE_DIR/etc/glance-api.conf
    cp $FILES/glance-api.conf $GLANCE_API_CONF
    sudo sed -e "s,%DEST%,$DEST,g" -i $GLANCE_API_CONF
    sudo sed -e "s,%SERVICE_TOKEN%,$SERVICE_TOKEN,g" -i $GLANCE_API_CONF
fi

# Nova
# ----

# We are going to use the sample http middleware configuration from the keystone
# project to launch nova.  This paste config adds the configuration required
# for nova to validate keystone tokens - except we need to switch the config
# to use our service token instead (instead of the invalid token 999888777666).
sudo sed -e "s,999888777666,$SERVICE_TOKEN,g" -i $KEYSTONE_DIR/examples/paste/nova-api-paste.ini

if [[ "$ENABLED_SERVICES" =~ "n-cpu" ]]; then

    # Virtualization Configuration
    # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~

    # attempt to load modules: network block device - used to manage qcow images
    sudo modprobe nbd || true

    # Check for kvm (hardware based virtualization).  If unable to initialize
    # kvm, we drop back to the slower emulation mode (qemu).  Note: many systems
    # come with hardware virtualization disabled in BIOS.
    if [[ "$LIBVIRT_TYPE" == "kvm" ]]; then
        sudo modprobe kvm || true
        if [ ! -e /dev/kvm ]; then
            echo "WARNING: Switching to QEMU"
            LIBVIRT_TYPE=qemu
        fi
    fi

    # Install and configure **LXC** if specified.  LXC is another approach to
    # splitting a system into many smaller parts.  LXC uses cgroups and chroot
    # to simulate multiple systems.
    if [[ "$LIBVIRT_TYPE" == "lxc" ]]; then
        apt_get install lxc
        # lxc uses cgroups (a kernel interface via virtual filesystem) configured
        # and mounted to ``/cgroup``
        sudo mkdir -p /cgroup
        if ! grep -q cgroup /etc/fstab; then
            echo none /cgroup cgroup cpuacct,memory,devices,cpu,freezer,blkio 0 0 | sudo tee -a /etc/fstab
        fi
        if ! mount -n | grep -q cgroup; then
            sudo mount /cgroup
        fi
    fi

    # The user that nova runs as needs to be member of libvirtd group otherwise
    # nova-compute will be unable to use libvirt.
    sudo usermod -a -G libvirtd `whoami`
    # libvirt detects various settings on startup, as we potentially changed
    # the system configuration (modules, filesystems), we need to restart
    # libvirt to detect those changes.
    sudo /etc/init.d/libvirt-bin restart


    # Instance Storage
    # ~~~~~~~~~~~~~~~~

    # Nova stores each instance in its own directory.
    mkdir -p $NOVA_DIR/instances

    # You can specify a different disk to be mounted and used for backing the
    # virtual machines.  If there is a partition labeled nova-instances we
    # mount it (ext filesystems can be labeled via e2label).
    if [ -L /dev/disk/by-label/nova-instances ]; then
        if ! mount -n | grep -q nova-instances; then
            sudo mount -L nova-instances $NOVA_DIR/instances
            sudo chown -R `whoami` $NOVA_DIR/instances
        fi
    fi

    # Clean out the instances directory.
    sudo rm -rf $NOVA_DIR/instances/*
fi

if [[ "$ENABLED_SERVICES" =~ "n-net" ]]; then
    # delete traces of nova networks from prior runs
    sudo killall dnsmasq || true
    rm -rf $NOVA_DIR/networks
    mkdir -p $NOVA_DIR/networks
fi

function add_nova_flag {
    echo "$1" >> $NOVA_DIR/bin/nova.conf
}

# (re)create nova.conf
rm -f $NOVA_DIR/bin/nova.conf
add_nova_flag "--verbose"
add_nova_flag "--nodaemon"
add_nova_flag "--scheduler_driver=$SCHEDULER"
add_nova_flag "--dhcpbridge_flagfile=$NOVA_DIR/bin/nova.conf"
add_nova_flag "--network_manager=nova.network.manager.$NET_MAN"
add_nova_flag "--my_ip=$HOST_IP"
add_nova_flag "--public_interface=$PUBLIC_INTERFACE"
add_nova_flag "--vlan_interface=$VLAN_INTERFACE"
add_nova_flag "--sql_connection=$BASE_SQL_CONN/nova"
add_nova_flag "--libvirt_type=$LIBVIRT_TYPE"
add_nova_flag "--osapi_extensions_path=$OPENSTACKX_DIR/extensions"
add_nova_flag "--vncproxy_url=http://$HOST_IP:6080"
add_nova_flag "--vncproxy_wwwroot=$NOVNC_DIR/"
add_nova_flag "--api_paste_config=$KEYSTONE_DIR/examples/paste/nova-api-paste.ini"
add_nova_flag "--image_service=nova.image.glance.GlanceImageService"
add_nova_flag "--ec2_dmz_host=$EC2_DMZ_HOST"
add_nova_flag "--rabbit_host=$RABBIT_HOST"
add_nova_flag "--rabbit_password=$RABBIT_PASSWORD"
add_nova_flag "--glance_api_servers=$GLANCE_HOSTPORT"
add_nova_flag "--flat_network_bridge=$FLAT_NETWORK_BRIDGE"
if [ -n "$FLAT_INTERFACE" ]; then
    add_nova_flag "--flat_interface=$FLAT_INTERFACE"
fi
if [ -n "$MULTI_HOST" ]; then
    add_nova_flag "--multi_host=$MULTI_HOST"
fi

# Nova Database
# ~~~~~~~~~~~~~

# All nova components talk to a central database.  We will need to do this step
# only once for an entire cluster.

if [[ "$ENABLED_SERVICES" =~ "mysql" ]]; then
    # (re)create nova database
    mysql -u$MYSQL_USER -p$MYSQL_PASSWORD -e 'DROP DATABASE IF EXISTS nova;'
    mysql -u$MYSQL_USER -p$MYSQL_PASSWORD -e 'CREATE DATABASE nova;'

    # (re)create nova database
    $NOVA_DIR/bin/nova-manage db sync

    # create a small network
    $NOVA_DIR/bin/nova-manage network create private $FIXED_RANGE 1 $FIXED_NETWORK_SIZE

    # create some floating ips
    $NOVA_DIR/bin/nova-manage floating create $FLOATING_RANGE
fi


# Keystone
# --------

if [[ "$ENABLED_SERVICES" =~ "key" ]]; then
    # (re)create keystone database
    mysql -u$MYSQL_USER -p$MYSQL_PASSWORD -e 'DROP DATABASE IF EXISTS keystone;'
    mysql -u$MYSQL_USER -p$MYSQL_PASSWORD -e 'CREATE DATABASE keystone;'

    # FIXME (anthony) keystone should use keystone.conf.example
    KEYSTONE_CONF=$KEYSTONE_DIR/etc/keystone.conf
    cp $FILES/keystone.conf $KEYSTONE_CONF
    sudo sed -e "s,%SQL_CONN%,$BASE_SQL_CONN/keystone,g" -i $KEYSTONE_CONF
    sudo sed -e "s,%DEST%,$DEST,g" -i $KEYSTONE_CONF

    # keystone_data.sh creates our admin user and our ``SERVICE_TOKEN``.
    KEYSTONE_DATA=$KEYSTONE_DIR/bin/keystone_data.sh
    cp $FILES/keystone_data.sh $KEYSTONE_DATA
    sudo sed -e "s,%HOST_IP%,$HOST_IP,g" -i $KEYSTONE_DATA
    sudo sed -e "s,%SERVICE_TOKEN%,$SERVICE_TOKEN,g" -i $KEYSTONE_DATA
    sudo sed -e "s,%ADMIN_PASSWORD%,$ADMIN_PASSWORD,g" -i $KEYSTONE_DATA
    # initialize keystone with default users/endpoints
    BIN_DIR=$KEYSTONE_DIR/bin bash $KEYSTONE_DATA
fi


# Launch Services
# ===============

# nova api crashes if we start it with a regular screen command,
# so send the start command by forcing text into the window.
# Only run the services specified in ``ENABLED_SERVICES``

# our screen helper to launch a service in a hidden named screen
function screen_it {
    NL=`echo -ne '\015'`
    if [[ "$ENABLED_SERVICES" =~ "$1" ]]; then
        screen -S nova -X screen -t $1
        screen -S nova -p $1 -X stuff "$2$NL"
    fi
}

# create a new named screen to run processes in
screen -d -m -S nova -t nova
sleep 1

# launch the glance registery service
if [[ "$ENABLED_SERVICES" =~ "g-reg" ]]; then
    screen_it g-reg "cd $GLANCE_DIR; bin/glance-registry --config-file=etc/glance-registry.conf"
fi

# launch the glance api and wait for it to answer before continuing
if [[ "$ENABLED_SERVICES" =~ "g-api" ]]; then
    screen_it g-api "cd $GLANCE_DIR; bin/glance-api --config-file=etc/glance-api.conf"
    while ! wget -q -O- http://$GLANCE_HOSTPORT; do
        echo "Waiting for g-api ($GLANCE_HOSTPORT) to start..."
        sleep 1
    done
fi

# launch the keystone and wait for it to answer before continuing
if [[ "$ENABLED_SERVICES" =~ "key" ]]; then
    screen_it key "cd $KEYSTONE_DIR && $KEYSTONE_DIR/bin/keystone --config-file $KEYSTONE_CONF -d"
    while ! wget -q -O- http://127.0.0.1:5000; do
        echo "Waiting for keystone to start..."
        sleep 1
    done
fi

# launch the nova-api and wait for it to answer before continuing
if [[ "$ENABLED_SERVICES" =~ "n-api" ]]; then
    screen_it n-api "cd $NOVA_DIR && $NOVA_DIR/bin/nova-api"
    while ! wget -q -O- http://127.0.0.1:8774; do
        echo "Waiting for nova-api to start..."
        sleep 1
    done
fi
# Launching nova-compute should be as simple as running ``nova-compute`` but
# have to do a little more than that in our script.  Since we add the group
# ``libvirtd`` to our user in this script, when nova-compute is run it is
# within the context of our original shell (so our groups won't be updated).
# Use 'sg' to execute nova-compute as a member of the libvirtd group.
screen_it n-cpu "cd $NOVA_DIR && sg libvirtd $NOVA_DIR/bin/nova-compute"
screen_it n-net "cd $NOVA_DIR && $NOVA_DIR/bin/nova-network"
screen_it n-sch "cd $NOVA_DIR && $NOVA_DIR/bin/nova-scheduler"
screen_it n-vnc "cd $NOVNC_DIR && ./utils/nova-wsproxy.py 6080 --web . --flagfile=../nova/bin/nova.conf"
screen_it dash "cd $DASH_DIR && sudo /etc/init.d/apache2 restart; sudo tail -f /var/log/apache2/error.log"

# Install Images
# ==============

# Upload an image to glance.
#
# The default image is a small ***TTY*** testing image, which lets you login
# the username/password of root/password.
#
# TTY also uses cloud-init, supporting login via keypair and sending scripts as
# userdata.  See https://help.ubuntu.com/community/CloudInit for more on cloud-init
#
# Override ``IMAGE_URLS`` with a comma-seperated list of uec images.
#
#  * **natty**: http://uec-images.ubuntu.com/natty/current/natty-server-cloudimg-amd64.tar.gz
#  * **oneiric**: http://uec-images.ubuntu.com/oneiric/current/oneiric-server-cloudimg-amd64.tar.gz

if [[ "$ENABLED_SERVICES" =~ "g-reg" ]]; then
    # Create a directory for the downloaded image tarballs.
    mkdir -p $FILES/images

    for image_url in ${IMAGE_URLS//,/ }; do
        # Downloads the image (uec ami+aki style), then extracts it.
        IMAGE_FNAME=`echo "$image_url" | python -c "import sys; print sys.stdin.read().split('/')[-1]"`
        IMAGE_NAME=`echo "$IMAGE_FNAME" | python -c "import sys; print sys.stdin.read().split('.tar.gz')[0].split('.tgz')[0]"`
        if [ ! -f $FILES/$IMAGE_FNAME ]; then
            wget -c $image_url -O $FILES/$IMAGE_FNAME
        fi

        # Extract ami and aki files
        tar -zxf $FILES/$IMAGE_FNAME -C $FILES/images

        # Use glance client to add the kernel the root filesystem.
        # We parse the results of the first upload to get the glance ID of the
        # kernel for use when uploading the root filesystem.
        RVAL=`glance add -A $SERVICE_TOKEN name="$IMAGE_NAME-kernel" is_public=true container_format=aki disk_format=aki < $FILES/images/$IMAGE_NAME-vmlinuz*`
        KERNEL_ID=`echo $RVAL | cut -d":" -f2 | tr -d " "`
        glance add -A $SERVICE_TOKEN name="$IMAGE_NAME" is_public=true container_format=ami disk_format=ami kernel_id=$KERNEL_ID < $FILES/images/$IMAGE_NAME.img
    done
fi

# Fin
# ===


) 2>&1 | tee "${LOGFILE}"

# Check that the left side of the above pipe succeeded
for ret in "${PIPESTATUS[@]}"; do [ $ret -eq 0 ] || exit $ret; done

(
# Using the cloud
# ===============

# If you installed the dashboard on this server, then you should be able
# to access the site using your browser.
if [[ "$ENABLED_SERVICES" =~ "dash" ]]; then
    echo "dashboard is now available at http://$HOST_IP/"
fi

# If keystone is present, you can point nova cli to this server
if [[ "$ENABLED_SERVICES" =~ "key" ]]; then
    echo "keystone is serving at http://$HOST_IP:5000/v2.0/"
    echo "examples on using novaclient command line is in exercise.sh"
    echo "the default users are: admin and demo"
    echo "the password: $ADMIN_PASSWORD"
fi

# indicate how long this took to run (bash maintained variable 'SECONDS')
echo "stack.sh completed in $SECONDS seconds."

) | tee -a "$LOGFILE"