#!/usr/bin/env bash # ``stack.sh`` is an opinionated OpenStack developer installation. It # installs and configures various combinations of **Ceilometer**, **Cinder**, # **Glance**, **Heat**, **Horizon**, **Keystone**, **Nova**, **Quantum** # and **Swift** # 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 a recent **Ubuntu** # (11.10 Oneiric or 12.04 Precise) or **Fedora** (F16 or F17) machine. It # should work in a VM or physical server. Additionally we put the list of # ``apt`` and ``rpm`` dependencies and other configuration files in this repo. # Learn more and get the most recent version at http://devstack.org # Keep track of the devstack directory TOP_DIR=$(cd $(dirname "$0") && pwd) # Import common functions source $TOP_DIR/functions # Determine what system we are running on. This provides ``os_VENDOR``, # ``os_RELEASE``, ``os_UPDATE``, ``os_PACKAGE``, ``os_CODENAME`` # and ``DISTRO`` GetDistro # Settings # ======== # ``stack.sh`` is customizable through setting environment variables. If you # want to override a setting you can set and export it:: # # 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:: # # MYSQL_PASSWORD=anothersecret # MYSQL_USER=hellaroot # # We try to have sensible defaults, so you should be able to run ``./stack.sh`` # in most cases. ``localrc`` is not distributed with DevStack and will never # be overwritten by a DevStack update. # # DevStack distributes ``stackrc`` which contains locations for the OpenStack # repositories and branches to configure. ``stackrc`` sources ``localrc`` to # allow you to safely override those settings. if [[ ! -r $TOP_DIR/stackrc ]]; then echo "ERROR: missing $TOP_DIR/stackrc - did you grab more than just stack.sh?" exit 1 fi source $TOP_DIR/stackrc # Proxy Settings # -------------- # HTTP and HTTPS proxy servers are supported via the usual environment variables [1] # ``http_proxy``, ``https_proxy`` and ``no_proxy``. They can be set in # ``localrc`` if necessary or on the command line:: # # [1] http://www.w3.org/Daemon/User/Proxies/ProxyClients.html # # http_proxy=http://proxy.example.com:3128/ no_proxy=repo.example.net ./stack.sh if [[ -n "$http_proxy" ]]; then export http_proxy=$http_proxy fi if [[ -n "$https_proxy" ]]; then export https_proxy=$https_proxy fi if [[ -n "$no_proxy" ]]; then export no_proxy=$no_proxy fi # Destination path for installation ``DEST`` DEST=${DEST:-/opt/stack} # Sanity Check # ============ # Remove services which were negated in ENABLED_SERVICES # using the "-" prefix (e.g., "-n-vol") instead of # calling disable_service(). disable_negated_services # Warn users who aren't on an explicitly supported distro, but allow them to # override check and attempt installation with ``FORCE=yes ./stack`` if [[ ! ${DISTRO} =~ (oneiric|precise|quantal|f16|f17) ]]; then echo "WARNING: this script has not been tested on $DISTRO" if [[ "$FORCE" != "yes" ]]; then echo "If you wish to run this script anyway run with FORCE=yes" exit 1 fi fi # Disallow qpid on oneiric if [ "${DISTRO}" = "oneiric" ] && is_service_enabled qpid ; then # Qpid was introduced in precise echo "You must use Ubuntu Precise or newer for Qpid support." exit 1 fi # ``stack.sh`` keeps function libraries here # Make sure ``$TOP_DIR/lib`` directory is present if [ ! -d $TOP_DIR/lib ]; then echo "ERROR: missing devstack/lib" exit 1 fi # ``stack.sh`` keeps the list of ``apt`` and ``rpm`` dependencies and config # templates and other useful files in the ``files`` subdirectory FILES=$TOP_DIR/files if [ ! -d $FILES ]; then echo "ERROR: missing devstack/files" exit 1 fi SCREEN_NAME=${SCREEN_NAME:-stack} # Check to see if we are already running DevStack if type -p screen >/dev/null && screen -ls | egrep -q "[0-9].$SCREEN_NAME"; then echo "You are already running a stack.sh session." echo "To rejoin this session type 'screen -x stack'." echo "To destroy this session, type './unstack.sh'." exit 1 fi # Make sure we only have one rpc backend enabled. rpc_backend_cnt=0 for svc in qpid zeromq rabbit; do is_service_enabled $svc && ((rpc_backend_cnt++)) done if [ "$rpc_backend_cnt" -gt 1 ]; then echo "ERROR: only one rpc backend may be enabled," echo " set only one of 'rabbit', 'qpid', 'zeromq'" echo " via ENABLED_SERVICES." elif [ "$rpc_backend_cnt" == 0 ]; then echo "ERROR: at least one rpc backend must be enabled," echo " set one of 'rabbit', 'qpid', 'zeromq'" echo " via ENABLED_SERVICES." fi unset rpc_backend_cnt # Make sure we only have one volume service enabled. if is_service_enabled cinder && is_service_enabled n-vol; then echo "ERROR: n-vol and cinder must not be enabled at the same time" exit 1 fi # Set up logging level VERBOSE=$(trueorfalse True $VERBOSE) # root Access # ----------- # OpenStack is designed to be run as a non-root user; Horizon will fail to run # as **root** since Apache will not 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 [[ $EUID -eq 0 ]]; then ROOTSLEEP=${ROOTSLEEP:-10} echo "You are running this script as root." echo "In $ROOTSLEEP seconds, we will create a user 'stack' and run as that user" sleep $ROOTSLEEP # Give the non-root user the ability to run as **root** via ``sudo`` is_package_installed sudo || install_package sudo if ! getent group stack >/dev/null; then echo "Creating a group called stack" groupadd stack fi if ! getent passwd stack >/dev/null; then echo "Creating a user called stack" useradd -g stack -s /bin/bash -d $DEST -m stack fi echo "Giving stack user passwordless sudo privileges" # UEC images ``/etc/sudoers`` does not have a ``#includedir``, add one grep -q "^#includedir.*/etc/sudoers.d" /etc/sudoers || echo "#includedir /etc/sudoers.d" >> /etc/sudoers ( umask 226 && echo "stack ALL=(ALL) NOPASSWD:ALL" \ > /etc/sudoers.d/50_stack_sh ) echo "Copying files to stack user" STACK_DIR="$DEST/${PWD##*/}" cp -r -f -T "$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 $STACK_DIR; bash stack.sh" stack fi exit 1 else # We're not **root**, make sure ``sudo`` is available is_package_installed sudo || die "Sudo is required. Re-run stack.sh as root ONE TIME ONLY to set up sudo." # UEC images ``/etc/sudoers`` does not have a ``#includedir``, add one sudo grep -q "^#includedir.*/etc/sudoers.d" /etc/sudoers || echo "#includedir /etc/sudoers.d" | sudo tee -a /etc/sudoers # Set up devstack sudoers TEMPFILE=`mktemp` echo "`whoami` ALL=(root) NOPASSWD:ALL" >$TEMPFILE # Some binaries might be under /sbin or /usr/sbin, so make sure sudo will # see them by forcing PATH echo "Defaults:`whoami` secure_path=/sbin:/usr/sbin:/usr/bin:/bin:/usr/local/sbin:/usr/local/bin" >> $TEMPFILE chmod 0440 $TEMPFILE sudo chown root:root $TEMPFILE sudo mv $TEMPFILE /etc/sudoers.d/50_stack_sh # Remove old file sudo rm -f /etc/sudoers.d/stack_sh_nova fi # 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 # Set ``OFFLINE`` to ``True`` to configure ``stack.sh`` to run cleanly without # Internet access. ``stack.sh`` must have been previously run with Internet # access to install prerequisites and fetch repositories. OFFLINE=`trueorfalse False $OFFLINE` # Set ``ERROR_ON_CLONE`` to ``True`` to configure ``stack.sh`` to exit if # the destination git repository does not exist during the ``git_clone`` # operation. ERROR_ON_CLONE=`trueorfalse False $ERROR_ON_CLONE` # Destination path for service data DATA_DIR=${DATA_DIR:-${DEST}/data} sudo mkdir -p $DATA_DIR sudo chown `whoami` $DATA_DIR # Common Configuration # ==================== # Set fixed and floating range here so we can make sure not to use addresses # from either range when attempting to guess the IP to use for the host. # Note that setting FIXED_RANGE may be necessary when running DevStack # in an OpenStack cloud that uses either of these address ranges internally. FLOATING_RANGE=${FLOATING_RANGE:-172.24.4.224/28} FIXED_RANGE=${FIXED_RANGE:-10.0.0.0/24} FIXED_NETWORK_SIZE=${FIXED_NETWORK_SIZE:-256} NETWORK_GATEWAY=${NETWORK_GATEWAY:-10.0.0.1} # Find the interface used for the default route HOST_IP_IFACE=${HOST_IP_IFACE:-$(ip route | sed -n '/^default/{ s/.*dev \(\w\+\)\s\+.*/\1/; p; }')} # Search for an IP unless an explicit is set by ``HOST_IP`` environment variable if [ -z "$HOST_IP" -o "$HOST_IP" == "dhcp" ]; then HOST_IP="" HOST_IPS=`LC_ALL=C ip -f inet addr show ${HOST_IP_IFACE} | awk '/inet/ {split($2,parts,"/"); print parts[1]}'` for IP in $HOST_IPS; do # Attempt to filter out IP addresses that are part of the fixed and # floating range. Note that this method only works if the ``netaddr`` # python library is installed. If it is not installed, an error # will be printed and the first IP from the interface will be used. # If that is not correct set ``HOST_IP`` in ``localrc`` to the correct # address. if ! (address_in_net $IP $FIXED_RANGE || address_in_net $IP $FLOATING_RANGE); then HOST_IP=$IP break; fi done if [ "$HOST_IP" == "" ]; then echo "Could not determine host ip address." echo "Either localrc specified dhcp on ${HOST_IP_IFACE} or defaulted" exit 1 fi fi # Allow the use of an alternate hostname (such as localhost/127.0.0.1) for service endpoints. SERVICE_HOST=${SERVICE_HOST:-$HOST_IP} # Configure services to use syslog instead of writing to individual log files SYSLOG=`trueorfalse False $SYSLOG` SYSLOG_HOST=${SYSLOG_HOST:-$HOST_IP} SYSLOG_PORT=${SYSLOG_PORT:-516} # Use color for logging output (only available if syslog is not used) LOG_COLOR=`trueorfalse True $LOG_COLOR` # Service startup timeout SERVICE_TIMEOUT=${SERVICE_TIMEOUT:-60} # Configure Projects # ================== # Get project function libraries source $TOP_DIR/lib/keystone source $TOP_DIR/lib/glance source $TOP_DIR/lib/nova source $TOP_DIR/lib/cinder source $TOP_DIR/lib/n-vol source $TOP_DIR/lib/ceilometer source $TOP_DIR/lib/heat source $TOP_DIR/lib/quantum source $TOP_DIR/lib/tempest # Set the destination directories for OpenStack projects HORIZON_DIR=$DEST/horizon OPENSTACKCLIENT_DIR=$DEST/python-openstackclient NOVNC_DIR=$DEST/noVNC SWIFT_DIR=$DEST/swift SWIFT3_DIR=$DEST/swift3 SWIFTCLIENT_DIR=$DEST/python-swiftclient QUANTUM_DIR=$DEST/quantum QUANTUM_CLIENT_DIR=$DEST/python-quantumclient # Default Quantum Plugin Q_PLUGIN=${Q_PLUGIN:-openvswitch} # Default Quantum Port Q_PORT=${Q_PORT:-9696} # Default Quantum Host Q_HOST=${Q_HOST:-localhost} # Which Quantum API nova should use # Default admin username Q_ADMIN_USERNAME=${Q_ADMIN_USERNAME:-quantum} # Default auth strategy Q_AUTH_STRATEGY=${Q_AUTH_STRATEGY:-keystone} # Use namespace or not Q_USE_NAMESPACE=${Q_USE_NAMESPACE:-True} Q_USE_ROOTWRAP=${Q_USE_ROOTWRAP=:-True} # Meta data IP Q_META_DATA_IP=${Q_META_DATA_IP:-$HOST_IP} # Name of the LVM volume group to use/create for iscsi volumes VOLUME_GROUP=${VOLUME_GROUP:-stack-volumes} VOLUME_NAME_PREFIX=${VOLUME_NAME_PREFIX:-volume-} INSTANCE_NAME_PREFIX=${INSTANCE_NAME_PREFIX:-instance-} # Generic helper to configure passwords function read_password { XTRACE=$(set +o | grep xtrace) 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 " echo "again. Use only alphanumeric characters." echo "If you leave this blank, a random default value will be used." pw=" " while true; do echo "Enter a password now:" read -e $var pw=${!var} [[ "$pw" = "`echo $pw | tr -cd [:alnum:]`" ]] && break echo "Invalid chars in password. Try again:" done if [ ! $pw ]; then pw=`openssl rand -hex 10` fi eval "$var=$pw" echo "$var=$pw" >> $localrc fi $XTRACE } # Nova Network Configuration # -------------------------- # FIXME: more documentation about why these are important options. Also # we should make sure we use the same variable names as the option names. if [ "$VIRT_DRIVER" = 'xenserver' ]; then PUBLIC_INTERFACE_DEFAULT=eth3 # Allow ``build_domU.sh`` to specify the flat network bridge via kernel args FLAT_NETWORK_BRIDGE_DEFAULT=$(grep -o 'flat_network_bridge=[[:alnum:]]*' /proc/cmdline | cut -d= -f 2 | sort -u) GUEST_INTERFACE_DEFAULT=eth1 else PUBLIC_INTERFACE_DEFAULT=br100 FLAT_NETWORK_BRIDGE_DEFAULT=br100 GUEST_INTERFACE_DEFAULT=eth0 fi PUBLIC_INTERFACE=${PUBLIC_INTERFACE:-$PUBLIC_INTERFACE_DEFAULT} NET_MAN=${NET_MAN:-FlatDHCPManager} EC2_DMZ_HOST=${EC2_DMZ_HOST:-$SERVICE_HOST} FLAT_NETWORK_BRIDGE=${FLAT_NETWORK_BRIDGE:-$FLAT_NETWORK_BRIDGE_DEFAULT} VLAN_INTERFACE=${VLAN_INTERFACE:-$GUEST_INTERFACE_DEFAULT} # Test floating pool and range are used for testing. They are defined # here until the admin APIs can replace nova-manage TEST_FLOATING_POOL=${TEST_FLOATING_POOL:-test} TEST_FLOATING_RANGE=${TEST_FLOATING_RANGE:-192.168.253.0/29} # ``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=`trueorfalse False $MULTI_HOST` # If you are using the FlatDHCP network mode on multiple hosts, set the # ``FLAT_INTERFACE`` variable but make sure that the interface doesn't already # have an IP or you risk breaking things. # # **DHCP Warning**: If your flat interface device uses DHCP, there will be a # hiccup while the network is moved from the flat interface to the flat network # bridge. This will happen when you launch your first instance. Upon launch # you will lose all connectivity to the node, and the VM launch will probably # fail. # # If you are running on a single node and don't need to access the VMs from # devices other than that node, you can set FLAT_INTERFACE= # This will stop nova from bridging any interfaces into FLAT_NETWORK_BRIDGE. FLAT_INTERFACE=${FLAT_INTERFACE-$GUEST_INTERFACE_DEFAULT} ## FIXME(ja): should/can we check that FLAT_INTERFACE is sane? # Using Quantum networking: # # Make sure that quantum is enabled in ENABLED_SERVICES. If you want # to run Quantum on this host, make sure that q-svc is also in # ENABLED_SERVICES. # # If you're planning to use the Quantum openvswitch plugin, set # Q_PLUGIN to "openvswitch" and make sure the q-agt service is enabled # in ENABLED_SERVICES. If you're planning to use the Quantum # linuxbridge plugin, set Q_PLUGIN to "linuxbridge" and make sure the # q-agt service is enabled in ENABLED_SERVICES. # # See "Quantum Network Configuration" below for additional variables # that must be set in localrc for connectivity across hosts with # Quantum. # # With Quantum networking the NET_MAN variable is ignored. # MySQL & (RabbitMQ or Qpid) # -------------------------- # We configure Nova, Horizon, Glance and Keystone to use MySQL as their # database server. While they share a single server, each has their own # database and tables. # By default this script will install and configure MySQL. If you want to # use an existing server, you can pass in the user/password/host parameters. # You will need to send the same ``MYSQL_PASSWORD`` to every host if you are doing # a multi-node DevStack installation. MYSQL_HOST=${MYSQL_HOST:-localhost} MYSQL_USER=${MYSQL_USER:-root} read_password MYSQL_PASSWORD "ENTER A PASSWORD TO USE FOR MYSQL." # NOTE: Don't specify ``/db`` in this string so we can use it for multiple services BASE_SQL_CONN=${BASE_SQL_CONN:-mysql://$MYSQL_USER:$MYSQL_PASSWORD@$MYSQL_HOST} # Rabbit connection info if is_service_enabled rabbit; then RABBIT_HOST=${RABBIT_HOST:-localhost} read_password RABBIT_PASSWORD "ENTER A PASSWORD TO USE FOR RABBIT." fi # Swift # ----- # TODO: add logging to different location. # Set ``SWIFT_DATA_DIR`` to the location of swift drives and objects. # Default is the common DevStack data directory. SWIFT_DATA_DIR=${SWIFT_DATA_DIR:-${DEST}/data/swift} # Set ``SWIFT_CONFIG_DIR`` to the location of the configuration files. # Default is ``/etc/swift``. SWIFT_CONFIG_DIR=${SWIFT_CONFIG_DIR:-/etc/swift} # DevStack will create a loop-back disk formatted as XFS to store the # swift data. Set ``SWIFT_LOOPBACK_DISK_SIZE`` to the disk size in bytes. # Default is 1 gigabyte. SWIFT_LOOPBACK_DISK_SIZE=${SWIFT_LOOPBACK_DISK_SIZE:-1000000} # The ring uses a configurable number of bits from a path’s MD5 hash as # a partition index that designates a device. The number of bits kept # from the hash is known as the partition power, and 2 to the partition # power indicates the partition count. Partitioning the full MD5 hash # ring allows other parts of the cluster to work in batches of items at # once which ends up either more efficient or at least less complex than # working with each item separately or the entire cluster all at once. # By default we define 9 for the partition count (which mean 512). SWIFT_PARTITION_POWER_SIZE=${SWIFT_PARTITION_POWER_SIZE:-9} # Set ``SWIFT_REPLICAS`` to configure how many replicas are to be # configured for your Swift cluster. By default the three replicas would need a # bit of IO and Memory on a VM you may want to lower that to 1 if you want to do # only some quick testing. SWIFT_REPLICAS=${SWIFT_REPLICAS:-3} if is_service_enabled swift; then # If we are using swift3, we can default the s3 port to swift instead # of nova-objectstore if is_service_enabled swift3;then S3_SERVICE_PORT=${S3_SERVICE_PORT:-8080} fi # We only ask for Swift Hash if we have enabled swift service. # SWIFT_HASH is a random unique string for a swift cluster that # can never change. read_password SWIFT_HASH "ENTER A RANDOM SWIFT HASH." fi # Set default port for nova-objectstore S3_SERVICE_PORT=${S3_SERVICE_PORT:-3333} # Keystone # -------- # The ``SERVICE_TOKEN`` is used to bootstrap the Keystone database. It is # just a string and is not a 'real' Keystone token. read_password SERVICE_TOKEN "ENTER A SERVICE_TOKEN TO USE FOR THE SERVICE ADMIN TOKEN." # Services authenticate to Identity with servicename/SERVICE_PASSWORD read_password SERVICE_PASSWORD "ENTER A SERVICE_PASSWORD TO USE FOR THE SERVICE AUTHENTICATION." # Horizon currently truncates usernames and passwords at 20 characters read_password ADMIN_PASSWORD "ENTER A PASSWORD TO USE FOR HORIZON AND KEYSTONE (20 CHARS OR LESS)." # Set the tenant for service accounts in Keystone SERVICE_TENANT_NAME=${SERVICE_TENANT_NAME:-service} # Horizon # ------- # Allow overriding the default Apache user and group, default both to # current user. APACHE_USER=${APACHE_USER:-$USER} APACHE_GROUP=${APACHE_GROUP:-$APACHE_USER} # Log files # --------- # Draw a spinner so the user knows something is happening function spinner() { local delay=0.75 local spinstr='|/-\' printf "..." >&3 while [ true ]; do local temp=${spinstr#?} printf "[%c]" "$spinstr" >&3 local spinstr=$temp${spinstr%"$temp"} sleep $delay printf "\b\b\b" >&3 done } # Echo text to the log file, summary log file and stdout # echo_summary "something to say" function echo_summary() { if [[ -t 3 && "$VERBOSE" != "True" ]]; then kill >/dev/null 2>&1 $LAST_SPINNER_PID if [ ! -z "$LAST_SPINNER_PID" ]; then printf "\b\b\bdone\n" >&3 fi echo -n $@ >&6 spinner & LAST_SPINNER_PID=$! else echo $@ >&6 fi } # Echo text only to stdout, no log files # echo_nolog "something not for the logs" function echo_nolog() { echo $@ >&3 } # Set up logging for ``stack.sh`` # Set ``LOGFILE`` to turn on logging # Append '.xxxxxxxx' to the given name to maintain history # where 'xxxxxxxx' is a representation of the date the file was created if [[ -n "$LOGFILE" || -n "$SCREEN_LOGDIR" ]]; then LOGDAYS=${LOGDAYS:-7} TIMESTAMP_FORMAT=${TIMESTAMP_FORMAT:-"%F-%H%M%S"} CURRENT_LOG_TIME=$(date "+$TIMESTAMP_FORMAT") fi if [[ -n "$LOGFILE" ]]; then # First clean up old log files. Use the user-specified ``LOGFILE`` # as the template to search for, appending '.*' to match the date # we added on earlier runs. LOGDIR=$(dirname "$LOGFILE") LOGNAME=$(basename "$LOGFILE") mkdir -p $LOGDIR find $LOGDIR -maxdepth 1 -name $LOGNAME.\* -mtime +$LOGDAYS -exec rm {} \; LOGFILE=$LOGFILE.${CURRENT_LOG_TIME} SUMFILE=$LOGFILE.${CURRENT_LOG_TIME}.summary # Redirect output according to config # Copy stdout to fd 3 exec 3>&1 if [[ "$VERBOSE" == "True" ]]; then # Redirect stdout/stderr to tee to write the log file exec 1> >( tee "${LOGFILE}" ) 2>&1 # Set up a second fd for output exec 6> >( tee "${SUMFILE}" ) else # Set fd 1 and 2 to primary logfile exec 1> "${LOGFILE}" 2>&1 # Set fd 6 to summary logfile and stdout exec 6> >( tee "${SUMFILE}" /dev/fd/3 ) fi echo_summary "stack.sh log $LOGFILE" # Specified logfile name always links to the most recent log ln -sf $LOGFILE $LOGDIR/$LOGNAME ln -sf $SUMFILE $LOGDIR/$LOGNAME.summary else # Set up output redirection without log files # Copy stdout to fd 3 exec 3>&1 if [[ "$VERBOSE" != "True" ]]; then # Throw away stdout and stderr exec 1>/dev/null 2>&1 fi # Always send summary fd to original stdout exec 6>&3 fi # Set up logging of screen windows # Set ``SCREEN_LOGDIR`` to turn on logging of screen windows to the # directory specified in ``SCREEN_LOGDIR``, we will log to the the file # ``screen-$SERVICE_NAME-$TIMESTAMP.log`` in that dir and have a link # ``screen-$SERVICE_NAME.log`` to the latest log file. # Logs are kept for as long specified in ``LOGDAYS``. if [[ -n "$SCREEN_LOGDIR" ]]; then # We make sure the directory is created. if [[ -d "$SCREEN_LOGDIR" ]]; then # We cleanup the old logs find $SCREEN_LOGDIR -maxdepth 1 -name screen-\*.log -mtime +$LOGDAYS -exec rm {} \; else mkdir -p $SCREEN_LOGDIR fi fi # Set Up Script Execution # ----------------------- # Kill background processes on exit trap clean EXIT clean() { local r=$? kill >/dev/null 2>&1 $(jobs -p) exit $r } # Exit on any errors so that errors don't compound trap failed ERR failed() { local r=$? kill >/dev/null 2>&1 $(jobs -p) set +o xtrace [ -n "$LOGFILE" ] && echo "${0##*/} failed: full log in $LOGFILE" exit $r } # Print the commands being run so that we can see the command that triggers # an error. It is also useful for following along as the install occurs. set -o xtrace # Install Packages # ================ # OpenStack uses a fair number of other projects. # Install package requirements echo_summary "Installing package prerequisites" if [[ "$os_PACKAGE" = "deb" ]]; then install_package $(get_packages $FILES/apts) else install_package $(get_packages $FILES/rpms) fi if [[ $SYSLOG != "False" ]]; then install_package rsyslog-relp fi if is_service_enabled rabbit; then # Install rabbitmq-server # the temp file is necessary due to LP: #878600 tfile=$(mktemp) install_package rabbitmq-server > "$tfile" 2>&1 cat "$tfile" rm -f "$tfile" elif is_service_enabled qpid; then if [[ "$os_PACKAGE" = "rpm" ]]; then install_package qpid-cpp-server-daemon else install_package qpidd fi elif is_service_enabled zeromq; then if [[ "$os_PACKAGE" = "rpm" ]]; then install_package zeromq python-zmq else install_package libzmq1 python-zmq fi fi if is_service_enabled mysql; then if [[ "$os_PACKAGE" = "deb" ]]; then # Seed configuration with mysql password so that apt-get install doesn't # prompt us for a password upon install. cat <$HOME/.my.cnf [client] user=$MYSQL_USER password=$MYSQL_PASSWORD host=$MYSQL_HOST EOF chmod 0600 $HOME/.my.cnf fi # Install mysql-server install_package mysql-server fi if is_service_enabled horizon; then if [[ "$os_PACKAGE" = "deb" ]]; then # Install apache2, which is NOPRIME'd install_package apache2 libapache2-mod-wsgi else sudo rm -f /etc/httpd/conf.d/000-* install_package httpd mod_wsgi fi fi if is_service_enabled q-agt; then if [[ "$Q_PLUGIN" = "openvswitch" ]]; then # Install deps # FIXME add to files/apts/quantum, but don't install if not needed! if [[ "$os_PACKAGE" = "deb" ]]; then kernel_version=`cat /proc/version | cut -d " " -f3` install_package make fakeroot dkms openvswitch-switch openvswitch-datapath-dkms linux-headers-$kernel_version else ### FIXME(dtroyer): Find RPMs for OpenVSwitch echo "OpenVSwitch packages need to be located" # Fedora does not started OVS by default restart_service openvswitch fi elif [[ "$Q_PLUGIN" = "linuxbridge" ]]; then install_package bridge-utils fi fi if is_service_enabled swift; then # Install memcached for swift. install_package memcached fi TRACK_DEPENDS=${TRACK_DEPENDS:-False} # Install python packages into a virtualenv so that we can track them if [[ $TRACK_DEPENDS = True ]] ; then echo_summary "Installing Python packages into a virtualenv $DEST/.venv" install_package python-virtualenv rm -rf $DEST/.venv virtualenv --system-site-packages $DEST/.venv source $DEST/.venv/bin/activate $DEST/.venv/bin/pip freeze > $DEST/requires-pre-pip fi # Install python requirements echo_summary "Installing Python prerequisites" pip_install $(get_packages $FILES/pips | sort -u) # Check Out Source # ---------------- echo_summary "Installing OpenStack project source" install_keystoneclient install_glanceclient install_novaclient # Check out the client libs that are used most git_clone $OPENSTACKCLIENT_REPO $OPENSTACKCLIENT_DIR $OPENSTACKCLIENT_BRANCH # glance, swift middleware and nova api needs keystone middleware if is_service_enabled key g-api n-api swift; then # unified auth system (manages accounts/tokens) install_keystone fi if is_service_enabled swift; then # storage service git_clone $SWIFT_REPO $SWIFT_DIR $SWIFT_BRANCH # storage service client and and Library git_clone $SWIFTCLIENT_REPO $SWIFTCLIENT_DIR $SWIFTCLIENT_BRANCH if is_service_enabled swift3; then # swift3 middleware to provide S3 emulation to Swift git_clone $SWIFT3_REPO $SWIFT3_DIR $SWIFT3_BRANCH fi fi if is_service_enabled g-api n-api; then # image catalog service install_glance fi if is_service_enabled nova; then # compute service install_nova fi if is_service_enabled n-novnc; then # a websockets/html5 or flash powered VNC console for vm instances git_clone $NOVNC_REPO $NOVNC_DIR $NOVNC_BRANCH fi if is_service_enabled horizon; then # django powered web control panel for openstack git_clone $HORIZON_REPO $HORIZON_DIR $HORIZON_BRANCH $HORIZON_TAG fi if is_service_enabled quantum; then git_clone $QUANTUM_CLIENT_REPO $QUANTUM_CLIENT_DIR $QUANTUM_CLIENT_BRANCH fi if is_service_enabled quantum; then # quantum git_clone $QUANTUM_REPO $QUANTUM_DIR $QUANTUM_BRANCH fi if is_service_enabled heat; then install_heat fi if is_service_enabled cinder; then install_cinder fi if is_service_enabled ceilometer; then install_ceilometer fi if is_service_enabled tempest; then install_tempest fi # Initialization # ============== echo_summary "Configuring OpenStack projects" # Set up our checkouts so they are installed into python path # allowing ``import nova`` or ``import glance.client`` configure_keystoneclient configure_novaclient setup_develop $OPENSTACKCLIENT_DIR if is_service_enabled key g-api n-api swift; then configure_keystone fi if is_service_enabled swift; then setup_develop $SWIFT_DIR setup_develop $SWIFTCLIENT_DIR fi if is_service_enabled swift3; then setup_develop $SWIFT3_DIR fi if is_service_enabled g-api n-api; then configure_glance fi # Do this _after_ glance is installed to override the old binary # TODO(dtroyer): figure out when this is no longer necessary configure_glanceclient if is_service_enabled nova; then configure_nova fi if is_service_enabled horizon; then setup_develop $HORIZON_DIR fi if is_service_enabled quantum; then setup_develop $QUANTUM_CLIENT_DIR setup_develop $QUANTUM_DIR fi if is_service_enabled heat; then configure_heat fi if is_service_enabled cinder; then configure_cinder fi if is_service_enabled tempest; then configure_tempest fi if [[ $TRACK_DEPENDS = True ]] ; then $DEST/.venv/bin/pip freeze > $DEST/requires-post-pip if ! diff -Nru $DEST/requires-pre-pip $DEST/requires-post-pip > $DEST/requires.diff ; then cat $DEST/requires.diff fi echo "Ran stack.sh in depend tracking mode, bailing out now" exit 0 fi # Syslog # ------ if [[ $SYSLOG != "False" ]]; then if [[ "$SYSLOG_HOST" = "$HOST_IP" ]]; then # Configure the master host to receive cat </tmp/90-stack-m.conf \$ModLoad imrelp \$InputRELPServerRun $SYSLOG_PORT EOF sudo mv /tmp/90-stack-m.conf /etc/rsyslog.d else # Set rsyslog to send to remote host cat </tmp/90-stack-s.conf *.* :omrelp:$SYSLOG_HOST:$SYSLOG_PORT EOF sudo mv /tmp/90-stack-s.conf /etc/rsyslog.d fi echo_summary "Starting rsyslog" restart_service rsyslog fi # Finalize queue installation # ---------------------------- if is_service_enabled rabbit; then # Start rabbitmq-server echo_summary "Starting RabbitMQ" if [[ "$os_PACKAGE" = "rpm" ]]; then # RPM doesn't start the service restart_service rabbitmq-server fi # change the rabbit password since the default is "guest" sudo rabbitmqctl change_password guest $RABBIT_PASSWORD elif is_service_enabled qpid; then echo_summary "Starting qpid" restart_service qpidd fi # Mysql # ----- if is_service_enabled mysql; then echo_summary "Configuring and starting MySQL" if [[ "$os_PACKAGE" = "deb" ]]; then MY_CONF=/etc/mysql/my.cnf MYSQL=mysql else MY_CONF=/etc/my.cnf MYSQL=mysqld fi # Start mysql-server if [[ "$os_PACKAGE" = "rpm" ]]; then # RPM doesn't start the service start_service $MYSQL # Set the root password - only works the first time sudo mysqladmin -u root password $MYSQL_PASSWORD || true fi # Update the DB to give user ‘$MYSQL_USER’@’%’ full control of the all databases: sudo mysql -uroot -p$MYSQL_PASSWORD -h127.0.0.1 -e "GRANT ALL PRIVILEGES ON *.* TO '$MYSQL_USER'@'%' identified by '$MYSQL_PASSWORD';" # Now update ``my.cnf`` for some local needs and restart the mysql service # Change ‘bind-address’ from localhost (127.0.0.1) to any (0.0.0.0) sudo sed -i '/^bind-address/s/127.0.0.1/0.0.0.0/g' $MY_CONF # Set default db type to InnoDB if sudo grep -q "default-storage-engine" $MY_CONF; then # Change it sudo bash -c "source $TOP_DIR/functions; iniset $MY_CONF mysqld default-storage-engine InnoDB" else # Add it sudo sed -i -e "/^\[mysqld\]/ a \ default-storage-engine = InnoDB" $MY_CONF fi restart_service $MYSQL fi if [ -z "$SCREEN_HARDSTATUS" ]; then SCREEN_HARDSTATUS='%{= .} %-Lw%{= .}%> %n%f %t*%{= .}%+Lw%< %-=%{g}(%{d}%H/%l%{g})' fi # Create a new named screen to run processes in screen -d -m -S $SCREEN_NAME -t shell -s /bin/bash sleep 1 # Set a reasonable status bar screen -r $SCREEN_NAME -X hardstatus alwayslastline "$SCREEN_HARDSTATUS" # Keystone # -------- if is_service_enabled key; then echo_summary "Starting Keystone" configure_keystone init_keystone start_keystone echo "Waiting for keystone to start..." if ! timeout $SERVICE_TIMEOUT sh -c "while ! http_proxy= curl -s $KEYSTONE_AUTH_PROTOCOL://$SERVICE_HOST:$KEYSTONE_API_PORT/v2.0/ >/dev/null; do sleep 1; done"; then echo "keystone did not start" exit 1 fi # ``keystone_data.sh`` creates services, admin and demo users, and roles. SERVICE_ENDPOINT=$KEYSTONE_AUTH_PROTOCOL://$KEYSTONE_AUTH_HOST:$KEYSTONE_AUTH_PORT/v2.0 ADMIN_PASSWORD=$ADMIN_PASSWORD SERVICE_TENANT_NAME=$SERVICE_TENANT_NAME SERVICE_PASSWORD=$SERVICE_PASSWORD \ SERVICE_TOKEN=$SERVICE_TOKEN SERVICE_ENDPOINT=$SERVICE_ENDPOINT SERVICE_HOST=$SERVICE_HOST \ S3_SERVICE_PORT=$S3_SERVICE_PORT KEYSTONE_CATALOG_BACKEND=$KEYSTONE_CATALOG_BACKEND \ DEVSTACK_DIR=$TOP_DIR ENABLED_SERVICES=$ENABLED_SERVICES HEAT_API_CFN_PORT=$HEAT_API_CFN_PORT \ HEAT_API_PORT=$HEAT_API_PORT \ bash -x $FILES/keystone_data.sh # Set up auth creds now that keystone is bootstrapped export OS_AUTH_URL=$SERVICE_ENDPOINT export OS_TENANT_NAME=admin export OS_USERNAME=admin export OS_PASSWORD=$ADMIN_PASSWORD fi # Horizon # ------- # Set up the django horizon application to serve via apache/wsgi if is_service_enabled horizon; then echo_summary "Configuring and starting Horizon" # Remove stale session database. rm -f $HORIZON_DIR/openstack_dashboard/local/dashboard_openstack.sqlite3 # ``local_settings.py`` is used to override horizon default settings. local_settings=$HORIZON_DIR/openstack_dashboard/local/local_settings.py cp $FILES/horizon_settings.py $local_settings # Initialize the horizon database (it stores sessions and notices shown to # users). The user system is external (keystone). cd $HORIZON_DIR python manage.py syncdb --noinput cd $TOP_DIR # Create an empty directory that apache uses as docroot sudo mkdir -p $HORIZON_DIR/.blackhole if [[ "$os_PACKAGE" = "deb" ]]; then APACHE_NAME=apache2 APACHE_CONF=sites-available/horizon # Clean up the old config name sudo rm -f /etc/apache2/sites-enabled/000-default # Be a good citizen and use the distro tools here sudo touch /etc/$APACHE_NAME/$APACHE_CONF sudo a2ensite horizon else # Install httpd, which is NOPRIME'd APACHE_NAME=httpd APACHE_CONF=conf.d/horizon.conf sudo sed '/^Listen/s/^.*$/Listen 0.0.0.0:80/' -i /etc/httpd/conf/httpd.conf fi # Configure apache to run horizon sudo sh -c "sed -e \" s,%USER%,$APACHE_USER,g; s,%GROUP%,$APACHE_GROUP,g; s,%HORIZON_DIR%,$HORIZON_DIR,g; s,%APACHE_NAME%,$APACHE_NAME,g; s,%DEST%,$DEST,g; \" $FILES/apache-horizon.template >/etc/$APACHE_NAME/$APACHE_CONF" restart_service $APACHE_NAME fi # Glance # ------ if is_service_enabled g-reg; then echo_summary "Configuring Glance" init_glance # Store the images in swift if enabled. if is_service_enabled swift; then iniset $GLANCE_API_CONF DEFAULT default_store swift iniset $GLANCE_API_CONF DEFAULT swift_store_auth_address $KEYSTONE_SERVICE_PROTOCOL://$KEYSTONE_SERVICE_HOST:$KEYSTONE_SERVICE_PORT/v2.0/ iniset $GLANCE_API_CONF DEFAULT swift_store_user $SERVICE_TENANT_NAME:glance iniset $GLANCE_API_CONF DEFAULT swift_store_key $SERVICE_PASSWORD iniset $GLANCE_API_CONF DEFAULT swift_store_create_container_on_put True fi fi # Quantum # ------- if is_service_enabled quantum; then echo_summary "Configuring Quantum" # # Quantum Network Configuration # # The following variables control the Quantum openvswitch and # linuxbridge plugins' allocation of tenant networks and # availability of provider networks. If these are not configured # in localrc, tenant networks will be local to the host (with no # remote connectivity), and no physical resources will be # available for the allocation of provider networks. # To use GRE tunnels for tenant networks, set to True in # localrc. GRE tunnels are only supported by the openvswitch # plugin, and currently only on Ubuntu. ENABLE_TENANT_TUNNELS=${ENABLE_TENANT_TUNNELS:-False} # If using GRE tunnels for tenant networks, specify the range of # tunnel IDs from which tenant networks are allocated. Can be # overriden in localrc in necesssary. TENANT_TUNNEL_RANGES=${TENANT_TUNNEL_RANGE:-1:1000} # To use VLANs for tenant networks, set to True in localrc. VLANs # are supported by the openvswitch and linuxbridge plugins, each # requiring additional configuration described below. ENABLE_TENANT_VLANS=${ENABLE_TENANT_VLANS:-False} # If using VLANs for tenant networks, set in localrc to specify # the range of VLAN VIDs from which tenant networks are # allocated. An external network switch must be configured to # trunk these VLANs between hosts for multi-host connectivity. # # Example: TENANT_VLAN_RANGE=1000:1999 TENANT_VLAN_RANGE=${TENANT_VLAN_RANGE:-} # If using VLANs for tenant networks, or if using flat or VLAN # provider networks, set in localrc to the name of the physical # network, and also configure OVS_PHYSICAL_BRIDGE for the # openvswitch agent or LB_PHYSICAL_INTERFACE for the linuxbridge # agent, as described below. # # Example: PHYSICAL_NETWORK=default PHYSICAL_NETWORK=${PHYSICAL_NETWORK:-} # With the openvswitch plugin, if using VLANs for tenant networks, # or if using flat or VLAN provider networks, set in localrc to # the name of the OVS bridge to use for the physical network. The # bridge will be created if it does not already exist, but a # physical interface must be manually added to the bridge as a # port for external connectivity. # # Example: OVS_PHYSICAL_BRIDGE=br-eth1 OVS_PHYSICAL_BRIDGE=${OVS_PHYSICAL_BRIDGE:-} # With the linuxbridge plugin, if using VLANs for tenant networks, # or if using flat or VLAN provider networks, set in localrc to # the name of the network interface to use for the physical # network. # # Example: LB_PHYSICAL_INTERFACE=eth1 LB_PHYSICAL_INTERFACE=${LB_PHYSICAL_INTERFACE:-} # With the openvswitch plugin, set to True in localrc to enable # provider GRE tunnels when ENABLE_TENANT_TUNNELS is False. # # Example: OVS_ENABLE_TUNNELING=True OVS_ENABLE_TUNNELING=${OVS_ENABLE_TUNNELING:-$ENABLE_TENANT_TUNNELS} # Put config files in ``/etc/quantum`` for everyone to find if [[ ! -d /etc/quantum ]]; then sudo mkdir -p /etc/quantum fi sudo chown `whoami` /etc/quantum if [[ "$Q_PLUGIN" = "openvswitch" ]]; then Q_PLUGIN_CONF_PATH=etc/quantum/plugins/openvswitch Q_PLUGIN_CONF_FILENAME=ovs_quantum_plugin.ini Q_DB_NAME="ovs_quantum" Q_PLUGIN_CLASS="quantum.plugins.openvswitch.ovs_quantum_plugin.OVSQuantumPluginV2" elif [[ "$Q_PLUGIN" = "linuxbridge" ]]; then Q_PLUGIN_CONF_PATH=etc/quantum/plugins/linuxbridge Q_PLUGIN_CONF_FILENAME=linuxbridge_conf.ini Q_DB_NAME="quantum_linux_bridge" Q_PLUGIN_CLASS="quantum.plugins.linuxbridge.lb_quantum_plugin.LinuxBridgePluginV2" else echo "Unknown Quantum plugin '$Q_PLUGIN'.. exiting" exit 1 fi # If needed, move config file from ``$QUANTUM_DIR/etc/quantum`` to ``/etc/quantum`` mkdir -p /$Q_PLUGIN_CONF_PATH Q_PLUGIN_CONF_FILE=$Q_PLUGIN_CONF_PATH/$Q_PLUGIN_CONF_FILENAME cp $QUANTUM_DIR/$Q_PLUGIN_CONF_FILE /$Q_PLUGIN_CONF_FILE iniset /$Q_PLUGIN_CONF_FILE DATABASE sql_connection mysql:\/\/$MYSQL_USER:$MYSQL_PASSWORD@$MYSQL_HOST\/$Q_DB_NAME?charset=utf8 Q_CONF_FILE=/etc/quantum/quantum.conf cp $QUANTUM_DIR/etc/quantum.conf $Q_CONF_FILE Q_RR_CONF_FILE=/etc/quantum/rootwrap.conf if [[ "$Q_USE_ROOTWRAP" == "False" ]]; then Q_RR_COMMAND="sudo" else Q_RR_COMMAND="sudo $QUANTUM_DIR/bin/quantum-rootwrap $Q_RR_CONF_FILE" fi cp -p $QUANTUM_DIR/etc/rootwrap.conf $Q_RR_CONF_FILE # Copy over the config and filter bits Q_CONF_ROOTWRAP_D=/etc/quantum/rootwrap.d mkdir -p $Q_CONF_ROOTWRAP_D cp -pr $QUANTUM_DIR/etc/quantum/rootwrap.d/* $Q_CONF_ROOTWRAP_D/ fi # Quantum service (for controller node) if is_service_enabled q-svc; then Q_API_PASTE_FILE=/etc/quantum/api-paste.ini Q_POLICY_FILE=/etc/quantum/policy.json cp $QUANTUM_DIR/etc/api-paste.ini $Q_API_PASTE_FILE cp $QUANTUM_DIR/etc/policy.json $Q_POLICY_FILE if is_service_enabled mysql; then mysql -u$MYSQL_USER -p$MYSQL_PASSWORD -e "DROP DATABASE IF EXISTS $Q_DB_NAME;" mysql -u$MYSQL_USER -p$MYSQL_PASSWORD -e "CREATE DATABASE IF NOT EXISTS $Q_DB_NAME CHARACTER SET utf8;" else echo "mysql must be enabled in order to use the $Q_PLUGIN Quantum plugin." exit 1 fi # Update either configuration file with plugin iniset $Q_CONF_FILE DEFAULT core_plugin $Q_PLUGIN_CLASS iniset $Q_CONF_FILE DEFAULT auth_strategy $Q_AUTH_STRATEGY quantum_setup_keystone $Q_API_PASTE_FILE filter:authtoken # Configure plugin if [[ "$Q_PLUGIN" = "openvswitch" ]]; then if [[ "$ENABLE_TENANT_TUNNELS" = "True" ]]; then iniset /$Q_PLUGIN_CONF_FILE OVS tenant_network_type gre iniset /$Q_PLUGIN_CONF_FILE OVS tunnel_id_ranges $TENANT_TUNNEL_RANGES elif [[ "$ENABLE_TENANT_VLANS" = "True" ]]; then iniset /$Q_PLUGIN_CONF_FILE OVS tenant_network_type vlan else echo "WARNING - The openvswitch plugin is using local tenant networks, with no connectivity between hosts." fi # Override OVS_VLAN_RANGES and OVS_BRIDGE_MAPPINGS in localrc # for more complex physical network configurations. if [[ "$OVS_VLAN_RANGES" = "" ]] && [[ "$PHYSICAL_NETWORK" != "" ]]; then OVS_VLAN_RANGES=$PHYSICAL_NETWORK if [[ "$TENANT_VLAN_RANGE" != "" ]]; then OVS_VLAN_RANGES=$OVS_VLAN_RANGES:$TENANT_VLAN_RANGE fi fi if [[ "$OVS_VLAN_RANGES" != "" ]]; then iniset /$Q_PLUGIN_CONF_FILE OVS network_vlan_ranges $OVS_VLAN_RANGES fi # Enable tunnel networks if selected if [[ $OVS_ENABLE_TUNNELING = "True" ]]; then iniset /$Q_PLUGIN_CONF_FILE OVS enable_tunneling True fi elif [[ "$Q_PLUGIN" = "linuxbridge" ]]; then if [[ "$ENABLE_TENANT_VLANS" = "True" ]]; then iniset /$Q_PLUGIN_CONF_FILE VLANS tenant_network_type vlan else echo "WARNING - The linuxbridge plugin is using local tenant networks, with no connectivity between hosts." fi # Override LB_VLAN_RANGES and LB_INTERFACE_MAPPINGS in localrc # for more complex physical network configurations. if [[ "$LB_VLAN_RANGES" = "" ]] && [[ "$PHYSICAL_NETWORK" != "" ]]; then LB_VLAN_RANGES=$PHYSICAL_NETWORK if [[ "$TENANT_VLAN_RANGE" != "" ]]; then LB_VLAN_RANGES=$LB_VLAN_RANGES:$TENANT_VLAN_RANGE fi fi if [[ "$LB_VLAN_RANGES" != "" ]]; then iniset /$Q_PLUGIN_CONF_FILE VLANS network_vlan_ranges $LB_VLAN_RANGES fi fi fi # Quantum agent (for compute nodes) if is_service_enabled q-agt; then # Configure agent for plugin if [[ "$Q_PLUGIN" = "openvswitch" ]]; then # Setup integration bridge OVS_BRIDGE=${OVS_BRIDGE:-br-int} quantum_setup_ovs_bridge $OVS_BRIDGE # Setup agent for tunneling if [[ "$OVS_ENABLE_TUNNELING" = "True" ]]; then # Verify tunnels are supported # REVISIT - also check kernel module support for GRE and patch ports OVS_VERSION=`ovs-vsctl --version | head -n 1 | awk '{print $4;}'` if [ $OVS_VERSION \< "1.4" ] && ! is_service_enabled q-svc ; then echo "You are running OVS version $OVS_VERSION." echo "OVS 1.4+ is required for tunneling between multiple hosts." exit 1 fi iniset /$Q_PLUGIN_CONF_FILE OVS enable_tunneling True iniset /$Q_PLUGIN_CONF_FILE OVS local_ip $HOST_IP fi # Setup physical network bridge mappings. Override # OVS_VLAN_RANGES and OVS_BRIDGE_MAPPINGS in localrc for more # complex physical network configurations. if [[ "$OVS_BRIDGE_MAPPINGS" = "" ]] && [[ "$PHYSICAL_NETWORK" != "" ]] && [[ "$OVS_PHYSICAL_BRIDGE" != "" ]]; then OVS_BRIDGE_MAPPINGS=$PHYSICAL_NETWORK:$OVS_PHYSICAL_BRIDGE # Configure bridge manually with physical interface as port for multi-node sudo ovs-vsctl --no-wait -- --may-exist add-br $OVS_PHYSICAL_BRIDGE fi if [[ "$OVS_BRIDGE_MAPPINGS" != "" ]]; then iniset /$Q_PLUGIN_CONF_FILE OVS bridge_mappings $OVS_BRIDGE_MAPPINGS fi AGENT_BINARY="$QUANTUM_DIR/bin/quantum-openvswitch-agent" elif [[ "$Q_PLUGIN" = "linuxbridge" ]]; then # Setup physical network interface mappings. Override # LB_VLAN_RANGES and LB_INTERFACE_MAPPINGS in localrc for more # complex physical network configurations. if [[ "$LB_INTERFACE_MAPPINGS" = "" ]] && [[ "$PHYSICAL_NETWORK" != "" ]] && [[ "$LB_PHYSICAL_INTERFACE" != "" ]]; then LB_INTERFACE_MAPPINGS=$PHYSICAL_NETWORK:$LB_PHYSICAL_INTERFACE fi if [[ "$LB_INTERFACE_MAPPINGS" != "" ]]; then iniset /$Q_PLUGIN_CONF_FILE LINUX_BRIDGE physical_interface_mappings $LB_INTERFACE_MAPPINGS fi AGENT_BINARY="$QUANTUM_DIR/bin/quantum-linuxbridge-agent" fi # Update config w/rootwrap iniset /$Q_PLUGIN_CONF_FILE AGENT root_helper "$Q_RR_COMMAND" fi # Quantum DHCP if is_service_enabled q-dhcp; then AGENT_DHCP_BINARY="$QUANTUM_DIR/bin/quantum-dhcp-agent" Q_DHCP_CONF_FILE=/etc/quantum/dhcp_agent.ini cp $QUANTUM_DIR/etc/dhcp_agent.ini $Q_DHCP_CONF_FILE # Set verbose iniset $Q_DHCP_CONF_FILE DEFAULT verbose True # Set debug iniset $Q_DHCP_CONF_FILE DEFAULT debug True iniset $Q_DHCP_CONF_FILE DEFAULT use_namespaces $Q_USE_NAMESPACE quantum_setup_keystone $Q_DHCP_CONF_FILE DEFAULT set_auth_url # Update config w/rootwrap iniset $Q_DHCP_CONF_FILE DEFAULT root_helper "$Q_RR_COMMAND" if [[ "$Q_PLUGIN" = "openvswitch" ]]; then iniset $Q_DHCP_CONF_FILE DEFAULT interface_driver quantum.agent.linux.interface.OVSInterfaceDriver elif [[ "$Q_PLUGIN" = "linuxbridge" ]]; then iniset $Q_DHCP_CONF_FILE DEFAULT interface_driver quantum.agent.linux.interface.BridgeInterfaceDriver fi fi # Quantum L3 if is_service_enabled q-l3; then AGENT_L3_BINARY="$QUANTUM_DIR/bin/quantum-l3-agent" PUBLIC_BRIDGE=${PUBLIC_BRIDGE:-br-ex} Q_L3_CONF_FILE=/etc/quantum/l3_agent.ini cp $QUANTUM_DIR/etc/l3_agent.ini $Q_L3_CONF_FILE # Set verbose iniset $Q_L3_CONF_FILE DEFAULT verbose True # Set debug iniset $Q_L3_CONF_FILE DEFAULT debug True iniset $Q_L3_CONF_FILE DEFAULT metadata_ip $Q_META_DATA_IP iniset $Q_L3_CONF_FILE DEFAULT use_namespaces $Q_USE_NAMESPACE iniset $Q_L3_CONF_FILE DEFAULT root_helper "$Q_RR_COMMAND" quantum_setup_keystone $Q_L3_CONF_FILE DEFAULT set_auth_url if [[ "$Q_PLUGIN" == "openvswitch" ]]; then iniset $Q_L3_CONF_FILE DEFAULT interface_driver quantum.agent.linux.interface.OVSInterfaceDriver iniset $Q_L3_CONF_FILE DEFAULT external_network_bridge $PUBLIC_BRIDGE # Set up external bridge # Create it if it does not exist sudo ovs-vsctl --no-wait -- --may-exist add-br $PUBLIC_BRIDGE # remove internal ports for PORT in `sudo ovs-vsctl --no-wait list-ports $PUBLIC_BRIDGE`; do TYPE=$(sudo ovs-vsctl get interface $PORT type) if [[ "$TYPE" == "internal" ]]; then echo `sudo ip link delete $PORT` > /dev/null sudo ovs-vsctl --no-wait del-port $bridge $PORT fi done # ensure no IP is configured on the public bridge sudo ip addr flush dev $PUBLIC_BRIDGE elif [[ "$Q_PLUGIN" = "linuxbridge" ]]; then iniset $Q_L3_CONF_FILE DEFAULT interface_driver quantum.agent.linux.interface.BridgeInterfaceDriver iniset $Q_L3_CONF_FILE DEFAULT external_network_bridge '' fi fi # Quantum RPC support - must be updated prior to starting any of the services if is_service_enabled quantum; then iniset $Q_CONF_FILE DEFAULT control_exchange quantum if is_service_enabled qpid ; then iniset $Q_CONF_FILE DEFAULT rpc_backend quantum.openstack.common.rpc.impl_qpid elif is_service_enabled zeromq; then iniset $Q_CONF_FILE DEFAULT rpc_backend quantum.openstack.common.rpc.impl_zmq elif [ -n "$RABBIT_HOST" ] && [ -n "$RABBIT_PASSWORD" ]; then iniset $Q_CONF_FILE DEFAULT rabbit_host $RABBIT_HOST iniset $Q_CONF_FILE DEFAULT rabbit_password $RABBIT_PASSWORD fi fi # Nova # ---- if is_service_enabled nova; then echo_summary "Configuring Nova" configure_nova fi if is_service_enabled n-net q-dhcp; then # Delete traces of nova networks from prior runs sudo killall dnsmasq || true clean_iptables rm -rf $NOVA_STATE_PATH/networks mkdir -p $NOVA_STATE_PATH/networks # Force IP forwarding on, just on case sudo sysctl -w net.ipv4.ip_forward=1 fi # Storage Service # --------------- if is_service_enabled swift; then echo_summary "Configuring Swift" # Make sure to kill all swift processes first swift-init all stop || true # First do a bit of setup by creating the directories and # changing the permissions so we can run it as our user. USER_GROUP=$(id -g) sudo mkdir -p ${SWIFT_DATA_DIR}/drives sudo chown -R $USER:${USER_GROUP} ${SWIFT_DATA_DIR} # Create a loopback disk and format it to XFS. if [[ -e ${SWIFT_DATA_DIR}/drives/images/swift.img ]]; then if egrep -q ${SWIFT_DATA_DIR}/drives/sdb1 /proc/mounts; then sudo umount ${SWIFT_DATA_DIR}/drives/sdb1 fi else mkdir -p ${SWIFT_DATA_DIR}/drives/images sudo touch ${SWIFT_DATA_DIR}/drives/images/swift.img sudo chown $USER: ${SWIFT_DATA_DIR}/drives/images/swift.img dd if=/dev/zero of=${SWIFT_DATA_DIR}/drives/images/swift.img \ bs=1024 count=0 seek=${SWIFT_LOOPBACK_DISK_SIZE} fi # Make a fresh XFS filesystem mkfs.xfs -f -i size=1024 ${SWIFT_DATA_DIR}/drives/images/swift.img # Mount the disk with mount options to make it as efficient as possible mkdir -p ${SWIFT_DATA_DIR}/drives/sdb1 if ! egrep -q ${SWIFT_DATA_DIR}/drives/sdb1 /proc/mounts; then sudo mount -t xfs -o loop,noatime,nodiratime,nobarrier,logbufs=8 \ ${SWIFT_DATA_DIR}/drives/images/swift.img ${SWIFT_DATA_DIR}/drives/sdb1 fi # Create a link to the above mount for x in $(seq ${SWIFT_REPLICAS}); do sudo ln -sf ${SWIFT_DATA_DIR}/drives/sdb1/$x ${SWIFT_DATA_DIR}/$x; done # Create all of the directories needed to emulate a few different servers for x in $(seq ${SWIFT_REPLICAS}); do drive=${SWIFT_DATA_DIR}/drives/sdb1/${x} node=${SWIFT_DATA_DIR}/${x}/node node_device=${node}/sdb1 [[ -d $node ]] && continue [[ -d $drive ]] && continue sudo install -o ${USER} -g $USER_GROUP -d $drive sudo install -o ${USER} -g $USER_GROUP -d $node_device sudo chown -R $USER: ${node} done sudo mkdir -p ${SWIFT_CONFIG_DIR}/{object,container,account}-server /var/run/swift sudo chown -R $USER: ${SWIFT_CONFIG_DIR} /var/run/swift if [[ "$SWIFT_CONFIG_DIR" != "/etc/swift" ]]; then # Some swift tools are hard-coded to use ``/etc/swift`` and are apparently not going to be fixed. # Create a symlink if the config dir is moved sudo ln -sf ${SWIFT_CONFIG_DIR} /etc/swift fi # Swift use rsync to synchronize between all the different # partitions (which make more sense when you have a multi-node # setup) we configure it with our version of rsync. sed -e " s/%GROUP%/${USER_GROUP}/; s/%USER%/$USER/; s,%SWIFT_DATA_DIR%,$SWIFT_DATA_DIR,; " $FILES/swift/rsyncd.conf | sudo tee /etc/rsyncd.conf if [[ "$os_PACKAGE" = "deb" ]]; then sudo sed -i '/^RSYNC_ENABLE=false/ { s/false/true/ }' /etc/default/rsync else sudo sed -i '/disable *= *yes/ { s/yes/no/ }' /etc/xinetd.d/rsync fi if is_service_enabled swift3;then swift_auth_server="s3token " fi # By default Swift will be installed with the tempauth middleware # which has some default username and password if you have # configured keystone it will checkout the directory. if is_service_enabled key; then swift_auth_server+="authtoken keystoneauth" else swift_auth_server=tempauth fi SWIFT_CONFIG_PROXY_SERVER=${SWIFT_CONFIG_DIR}/proxy-server.conf cp ${SWIFT_DIR}/etc/proxy-server.conf-sample ${SWIFT_CONFIG_PROXY_SERVER} iniuncomment ${SWIFT_CONFIG_PROXY_SERVER} DEFAULT user iniset ${SWIFT_CONFIG_PROXY_SERVER} DEFAULT user ${USER} iniuncomment ${SWIFT_CONFIG_PROXY_SERVER} DEFAULT swift_dir iniset ${SWIFT_CONFIG_PROXY_SERVER} DEFAULT swift_dir ${SWIFT_CONFIG_DIR} iniuncomment ${SWIFT_CONFIG_PROXY_SERVER} DEFAULT workers iniset ${SWIFT_CONFIG_PROXY_SERVER} DEFAULT workers 1 iniuncomment ${SWIFT_CONFIG_PROXY_SERVER} DEFAULT log_level iniset ${SWIFT_CONFIG_PROXY_SERVER} DEFAULT log_level DEBUG iniuncomment ${SWIFT_CONFIG_PROXY_SERVER} DEFAULT bind_port iniset ${SWIFT_CONFIG_PROXY_SERVER} DEFAULT bind_port ${SWIFT_DEFAULT_BIND_PORT:-8080} # Only enable Swift3 if we have it enabled in ENABLED_SERVICES is_service_enabled swift3 && swift3=swift3 || swift3="" iniset ${SWIFT_CONFIG_PROXY_SERVER} pipeline:main pipeline "catch_errors healthcheck cache ratelimit ${swift3} ${swift_auth_server} proxy-logging proxy-server" iniset ${SWIFT_CONFIG_PROXY_SERVER} app:proxy-server account_autocreate true # Configure Keystone sed -i '/^# \[filter:authtoken\]/,/^# \[filter:keystoneauth\]$/ s/^#[ \t]*//' ${SWIFT_CONFIG_PROXY_SERVER} iniset ${SWIFT_CONFIG_PROXY_SERVER} filter:authtoken auth_host $KEYSTONE_AUTH_HOST iniset ${SWIFT_CONFIG_PROXY_SERVER} filter:authtoken auth_port $KEYSTONE_AUTH_PORT iniset ${SWIFT_CONFIG_PROXY_SERVER} filter:authtoken auth_protocol $KEYSTONE_AUTH_PROTOCOL iniset ${SWIFT_CONFIG_PROXY_SERVER} filter:authtoken auth_uri $KEYSTONE_SERVICE_PROTOCOL://$KEYSTONE_SERVICE_HOST:$KEYSTONE_SERVICE_PORT/ iniset ${SWIFT_CONFIG_PROXY_SERVER} filter:authtoken admin_tenant_name $SERVICE_TENANT_NAME iniset ${SWIFT_CONFIG_PROXY_SERVER} filter:authtoken admin_user swift iniset ${SWIFT_CONFIG_PROXY_SERVER} filter:authtoken admin_password $SERVICE_PASSWORD iniuncomment ${SWIFT_CONFIG_PROXY_SERVER} filter:keystoneauth use iniuncomment ${SWIFT_CONFIG_PROXY_SERVER} filter:keystoneauth operator_roles iniset ${SWIFT_CONFIG_PROXY_SERVER} filter:keystoneauth operator_roles "Member, admin" if is_service_enabled swift3;then cat <>${SWIFT_CONFIG_PROXY_SERVER} # NOTE(chmou): s3token middleware is not updated yet to use only # username and password. [filter:s3token] paste.filter_factory = keystone.middleware.s3_token:filter_factory auth_port = ${KEYSTONE_AUTH_PORT} auth_host = ${KEYSTONE_AUTH_HOST} auth_protocol = ${KEYSTONE_AUTH_PROTOCOL} auth_token = ${SERVICE_TOKEN} admin_token = ${SERVICE_TOKEN} [filter:swift3] use = egg:swift3#swift3 EOF fi cp ${SWIFT_DIR}/etc/swift.conf-sample ${SWIFT_CONFIG_DIR}/swift.conf iniset ${SWIFT_CONFIG_DIR}/swift.conf swift-hash swift_hash_path_suffix ${SWIFT_HASH} # This function generates an object/account/proxy configuration # emulating 4 nodes on different ports function generate_swift_configuration() { local server_type=$1 local bind_port=$2 local log_facility=$3 local node_number local swift_node_config for node_number in $(seq ${SWIFT_REPLICAS}); do node_path=${SWIFT_DATA_DIR}/${node_number} swift_node_config=${SWIFT_CONFIG_DIR}/${server_type}-server/${node_number}.conf cp ${SWIFT_DIR}/etc/${server_type}-server.conf-sample ${swift_node_config} iniuncomment ${swift_node_config} DEFAULT user iniset ${swift_node_config} DEFAULT user ${USER} iniuncomment ${swift_node_config} DEFAULT bind_port iniset ${swift_node_config} DEFAULT bind_port ${bind_port} iniuncomment ${swift_node_config} DEFAULT swift_dir iniset ${swift_node_config} DEFAULT swift_dir ${SWIFT_CONFIG_DIR} iniuncomment ${swift_node_config} DEFAULT devices iniset ${swift_node_config} DEFAULT devices ${node_path} iniuncomment ${swift_node_config} DEFAULT log_facility iniset ${swift_node_config} DEFAULT log_facility LOG_LOCAL${log_facility} iniuncomment ${swift_node_config} DEFAULT mount_check iniset ${swift_node_config} DEFAULT mount_check false iniuncomment ${swift_node_config} ${server_type}-replicator vm_test_mode iniset ${swift_node_config} ${server_type}-replicator vm_test_mode yes bind_port=$(( ${bind_port} + 10 )) log_facility=$(( ${log_facility} + 1 )) done } generate_swift_configuration object 6010 2 generate_swift_configuration container 6011 2 generate_swift_configuration account 6012 2 # Specific configuration for swift for rsyslog. See # ``/etc/rsyslog.d/10-swift.conf`` for more info. swift_log_dir=${SWIFT_DATA_DIR}/logs rm -rf ${swift_log_dir} mkdir -p ${swift_log_dir}/hourly sudo chown -R $USER:adm ${swift_log_dir} sed "s,%SWIFT_LOGDIR%,${swift_log_dir}," $FILES/swift/rsyslog.conf | sudo \ tee /etc/rsyslog.d/10-swift.conf restart_service rsyslog # This is where we create three different rings for swift with # different object servers binding on different ports. pushd ${SWIFT_CONFIG_DIR} >/dev/null && { rm -f *.builder *.ring.gz backups/*.builder backups/*.ring.gz port_number=6010 swift-ring-builder object.builder create ${SWIFT_PARTITION_POWER_SIZE} ${SWIFT_REPLICAS} 1 for x in $(seq ${SWIFT_REPLICAS}); do swift-ring-builder object.builder add z${x}-127.0.0.1:${port_number}/sdb1 1 port_number=$[port_number + 10] done swift-ring-builder object.builder rebalance port_number=6011 swift-ring-builder container.builder create ${SWIFT_PARTITION_POWER_SIZE} ${SWIFT_REPLICAS} 1 for x in $(seq ${SWIFT_REPLICAS}); do swift-ring-builder container.builder add z${x}-127.0.0.1:${port_number}/sdb1 1 port_number=$[port_number + 10] done swift-ring-builder container.builder rebalance port_number=6012 swift-ring-builder account.builder create ${SWIFT_PARTITION_POWER_SIZE} ${SWIFT_REPLICAS} 1 for x in $(seq ${SWIFT_REPLICAS}); do swift-ring-builder account.builder add z${x}-127.0.0.1:${port_number}/sdb1 1 port_number=$[port_number + 10] done swift-ring-builder account.builder rebalance } && popd >/dev/null # Start rsync if [[ "$os_PACKAGE" = "deb" ]]; then sudo /etc/init.d/rsync restart || : else sudo systemctl start xinetd.service fi # First spawn all the swift services then kill the # proxy service so we can run it in foreground in screen. # ``swift-init ... {stop|restart}`` exits with '1' if no servers are running, # ignore it just in case swift-init all restart || true swift-init proxy stop || true unset s swift_hash swift_auth_server fi # Volume Service # -------------- if is_service_enabled cinder; then echo_summary "Configuring Cinder" init_cinder elif is_service_enabled n-vol; then echo_summary "Configuring Nova volumes" init_nvol fi if is_service_enabled nova; then echo_summary "Configuring Nova" init_nova fi # Additional Nova configuration that is dependent on other services if is_service_enabled quantum; then add_nova_opt "network_api_class=nova.network.quantumv2.api.API" add_nova_opt "quantum_admin_username=$Q_ADMIN_USERNAME" add_nova_opt "quantum_admin_password=$SERVICE_PASSWORD" add_nova_opt "quantum_admin_auth_url=$KEYSTONE_SERVICE_PROTOCOL://$KEYSTONE_SERVICE_HOST:$KEYSTONE_AUTH_PORT/v2.0" add_nova_opt "quantum_auth_strategy=$Q_AUTH_STRATEGY" add_nova_opt "quantum_admin_tenant_name=$SERVICE_TENANT_NAME" add_nova_opt "quantum_url=http://$Q_HOST:$Q_PORT" if [[ "$Q_PLUGIN" = "openvswitch" ]]; then NOVA_VIF_DRIVER="nova.virt.libvirt.vif.LibvirtHybridOVSBridgeDriver" elif [[ "$Q_PLUGIN" = "linuxbridge" ]]; then NOVA_VIF_DRIVER="nova.virt.libvirt.vif.QuantumLinuxBridgeVIFDriver" fi add_nova_opt "libvirt_vif_driver=$NOVA_VIF_DRIVER" add_nova_opt "linuxnet_interface_driver=$LINUXNET_VIF_DRIVER" else add_nova_opt "network_manager=nova.network.manager.$NET_MAN" add_nova_opt "public_interface=$PUBLIC_INTERFACE" add_nova_opt "vlan_interface=$VLAN_INTERFACE" add_nova_opt "flat_network_bridge=$FLAT_NETWORK_BRIDGE" if [ -n "$FLAT_INTERFACE" ]; then add_nova_opt "flat_interface=$FLAT_INTERFACE" fi fi # All nova-compute workers need to know the vnc configuration options # These settings don't hurt anything if n-xvnc and n-novnc are disabled if is_service_enabled n-cpu; then NOVNCPROXY_URL=${NOVNCPROXY_URL:-"http://$SERVICE_HOST:6080/vnc_auto.html"} add_nova_opt "novncproxy_base_url=$NOVNCPROXY_URL" XVPVNCPROXY_URL=${XVPVNCPROXY_URL:-"http://$SERVICE_HOST:6081/console"} add_nova_opt "xvpvncproxy_base_url=$XVPVNCPROXY_URL" fi if [ "$VIRT_DRIVER" = 'xenserver' ]; then VNCSERVER_PROXYCLIENT_ADDRESS=${VNCSERVER_PROXYCLIENT_ADDRESS=169.254.0.1} else VNCSERVER_PROXYCLIENT_ADDRESS=${VNCSERVER_PROXYCLIENT_ADDRESS=127.0.0.1} fi # Address on which instance vncservers will listen on compute hosts. # For multi-host, this should be the management ip of the compute host. VNCSERVER_LISTEN=${VNCSERVER_LISTEN=127.0.0.1} add_nova_opt "vncserver_listen=$VNCSERVER_LISTEN" add_nova_opt "vncserver_proxyclient_address=$VNCSERVER_PROXYCLIENT_ADDRESS" add_nova_opt "ec2_dmz_host=$EC2_DMZ_HOST" if is_service_enabled zeromq; then add_nova_opt "rpc_backend=nova.openstack.common.rpc.impl_zmq" elif is_service_enabled qpid; then add_nova_opt "rpc_backend=nova.rpc.impl_qpid" elif [ -n "$RABBIT_HOST" ] && [ -n "$RABBIT_PASSWORD" ]; then add_nova_opt "rabbit_host=$RABBIT_HOST" add_nova_opt "rabbit_password=$RABBIT_PASSWORD" fi add_nova_opt "glance_api_servers=$GLANCE_HOSTPORT" # XenServer # --------- if [ "$VIRT_DRIVER" = 'xenserver' ]; then echo_summary "Using XenServer virtualization driver" read_password XENAPI_PASSWORD "ENTER A PASSWORD TO USE FOR XEN." add_nova_opt "compute_driver=xenapi.XenAPIDriver" XENAPI_CONNECTION_URL=${XENAPI_CONNECTION_URL:-"http://169.254.0.1"} XENAPI_USER=${XENAPI_USER:-"root"} add_nova_opt "xenapi_connection_url=$XENAPI_CONNECTION_URL" add_nova_opt "xenapi_connection_username=$XENAPI_USER" add_nova_opt "xenapi_connection_password=$XENAPI_PASSWORD" add_nova_opt "flat_injected=False" # Need to avoid crash due to new firewall support XEN_FIREWALL_DRIVER=${XEN_FIREWALL_DRIVER:-"nova.virt.firewall.IptablesFirewallDriver"} add_nova_opt "firewall_driver=$XEN_FIREWALL_DRIVER" elif [ "$VIRT_DRIVER" = 'openvz' ]; then echo_summary "Using OpenVZ virtualization driver" # TODO(deva): OpenVZ driver does not yet work if compute_driver is set here. # Replace connection_type when this is fixed. # add_nova_opt "compute_driver=openvz.connection.OpenVzConnection" add_nova_opt "connection_type=openvz" LIBVIRT_FIREWALL_DRIVER=${LIBVIRT_FIREWALL_DRIVER:-"nova.virt.libvirt.firewall.IptablesFirewallDriver"} add_nova_opt "firewall_driver=$LIBVIRT_FIREWALL_DRIVER" else echo_summary "Using libvirt virtualization driver" add_nova_opt "compute_driver=libvirt.LibvirtDriver" LIBVIRT_FIREWALL_DRIVER=${LIBVIRT_FIREWALL_DRIVER:-"nova.virt.libvirt.firewall.IptablesFirewallDriver"} add_nova_opt "firewall_driver=$LIBVIRT_FIREWALL_DRIVER" fi # Heat # ---- if is_service_enabled heat; then echo_summary "Configuring Heat" init_heat fi # Launch Services # =============== # Only run the services specified in ``ENABLED_SERVICES`` # Launch the Glance services if is_service_enabled g-api g-reg; then echo_summary "Starting Glance" start_glance fi # Create an access key and secret key for nova ec2 register image if is_service_enabled key && is_service_enabled swift3 && is_service_enabled nova; then NOVA_USER_ID=$(keystone user-list | grep ' nova ' | get_field 1) NOVA_TENANT_ID=$(keystone tenant-list | grep " $SERVICE_TENANT_NAME " | get_field 1) CREDS=$(keystone ec2-credentials-create --user_id $NOVA_USER_ID --tenant_id $NOVA_TENANT_ID) ACCESS_KEY=$(echo "$CREDS" | awk '/ access / { print $4 }') SECRET_KEY=$(echo "$CREDS" | awk '/ secret / { print $4 }') add_nova_opt "s3_access_key=$ACCESS_KEY" add_nova_opt "s3_secret_key=$SECRET_KEY" add_nova_opt "s3_affix_tenant=True" fi screen_it zeromq "cd $NOVA_DIR && $NOVA_DIR/bin/nova-rpc-zmq-receiver" # Launch the nova-api and wait for it to answer before continuing if is_service_enabled n-api; then echo_summary "Starting Nova API" screen_it n-api "cd $NOVA_DIR && $NOVA_BIN_DIR/nova-api" echo "Waiting for nova-api to start..." if ! timeout $SERVICE_TIMEOUT sh -c "while ! http_proxy= wget -q -O- http://127.0.0.1:8774; do sleep 1; done"; then echo "nova-api did not start" exit 1 fi fi if is_service_enabled q-svc; then echo_summary "Starting Quantum" # Start the Quantum service screen_it q-svc "cd $QUANTUM_DIR && python $QUANTUM_DIR/bin/quantum-server --config-file $Q_CONF_FILE --config-file /$Q_PLUGIN_CONF_FILE" echo "Waiting for Quantum to start..." if ! timeout $SERVICE_TIMEOUT sh -c "while ! http_proxy= wget -q -O- http://127.0.0.1:9696; do sleep 1; done"; then echo "Quantum did not start" exit 1 fi # Configure Quantum elements # Configure internal network & subnet TENANT_ID=$(keystone tenant-list | grep " demo " | get_field 1) # Create a small network # Since quantum command is executed in admin context at this point, # ``--tenant_id`` needs to be specified. NET_ID=$(quantum net-create --tenant_id $TENANT_ID net1 | grep ' id ' | get_field 2) SUBNET_ID=$(quantum subnet-create --tenant_id $TENANT_ID --ip_version 4 --gateway $NETWORK_GATEWAY $NET_ID $FIXED_RANGE | grep ' id ' | get_field 2) if is_service_enabled q-l3; then # Create a router, and add the private subnet as one of its interfaces ROUTER_ID=$(quantum router-create --tenant_id $TENANT_ID router1 | grep ' id ' | get_field 2) quantum router-interface-add $ROUTER_ID $SUBNET_ID # Create an external network, and a subnet. Configure the external network as router gw EXT_NET_ID=$(quantum net-create ext_net -- --router:external=True | grep ' id ' | get_field 2) EXT_GW_IP=$(quantum subnet-create --ip_version 4 $EXT_NET_ID $FLOATING_RANGE -- --enable_dhcp=False | grep 'gateway_ip' | get_field 2) quantum router-gateway-set $ROUTER_ID $EXT_NET_ID if [[ "$Q_PLUGIN" = "openvswitch" ]] && [[ "$Q_USE_NAMESPACE" = "True" ]]; then CIDR_LEN=${FLOATING_RANGE#*/} sudo ip addr add $EXT_GW_IP/$CIDR_LEN dev $PUBLIC_BRIDGE sudo ip link set $PUBLIC_BRIDGE up ROUTER_GW_IP=`quantum port-list -c fixed_ips -c device_owner | grep router_gateway | awk -F '"' '{ print $8; }'` sudo route add -net $FIXED_RANGE gw $ROUTER_GW_IP fi if [[ "$Q_USE_NAMESPACE" == "False" ]]; then # Explicitly set router id in l3 agent configuration iniset $Q_L3_CONF_FILE DEFAULT router_id $ROUTER_ID fi fi elif is_service_enabled mysql && is_service_enabled nova; then # Create a small network $NOVA_BIN_DIR/nova-manage network create private $FIXED_RANGE 1 $FIXED_NETWORK_SIZE $NETWORK_CREATE_ARGS # Create some floating ips $NOVA_BIN_DIR/nova-manage floating create $FLOATING_RANGE # Create a second pool $NOVA_BIN_DIR/nova-manage floating create --ip_range=$TEST_FLOATING_RANGE --pool=$TEST_FLOATING_POOL fi # Start up the quantum agents if enabled screen_it q-agt "python $AGENT_BINARY --config-file $Q_CONF_FILE --config-file /$Q_PLUGIN_CONF_FILE" screen_it q-dhcp "python $AGENT_DHCP_BINARY --config-file $Q_CONF_FILE --config-file=$Q_DHCP_CONF_FILE" screen_it q-l3 "python $AGENT_L3_BINARY --config-file $Q_CONF_FILE --config-file=$Q_L3_CONF_FILE" if is_service_enabled nova; then echo_summary "Starting Nova" start_nova fi if is_service_enabled n-vol; then echo_summary "Starting Nova volumes" start_nvol fi if is_service_enabled cinder; then echo_summary "Starting Cinder" start_cinder fi if is_service_enabled ceilometer; then echo_summary "Configuring Ceilometer" configure_ceilometer echo_summary "Starting Ceilometer" start_ceilometer fi screen_it horizon "cd $HORIZON_DIR && sudo tail -f /var/log/$APACHE_NAME/horizon_error.log" screen_it swift "cd $SWIFT_DIR && $SWIFT_DIR/bin/swift-proxy-server ${SWIFT_CONFIG_DIR}/proxy-server.conf -v" # Starting the nova-objectstore only if swift3 service is not enabled. # Swift will act as s3 objectstore. is_service_enabled swift3 || \ screen_it n-obj "cd $NOVA_DIR && $NOVA_BIN_DIR/nova-objectstore" # launch heat engine, api and metadata if is_service_enabled heat; then echo_summary "Starting Heat" start_heat fi # Install Images # ============== # Upload an image to glance. # # The default image is cirros, a small testing image which lets you login as **root** # cirros also uses ``cloud-init``, supporting login via keypair and sending scripts as # userdata. See https://help.ubuntu.com/community/CloudInit for more on cloud-init # # Override ``IMAGE_URLS`` with a comma-separated list of UEC images. # * **oneiric**: http://uec-images.ubuntu.com/oneiric/current/oneiric-server-cloudimg-amd64.tar.gz # * **precise**: http://uec-images.ubuntu.com/precise/current/precise-server-cloudimg-amd64.tar.gz if is_service_enabled g-reg; then echo_summary "Uploading images" TOKEN=$(keystone token-get | grep ' id ' | get_field 2) # Option to upload legacy ami-tty, which works with xenserver if [[ -n "$UPLOAD_LEGACY_TTY" ]]; then IMAGE_URLS="${IMAGE_URLS:+${IMAGE_URLS},}http://images.ansolabs.com/tty.tgz" fi for image_url in ${IMAGE_URLS//,/ }; do upload_image $image_url $TOKEN done fi # Run local script # ================ # Run ``local.sh`` if it exists to perform user-managed tasks if [[ -x $TOP_DIR/local.sh ]]; then echo "Running user script $TOP_DIR/local.sh" $TOP_DIR/local.sh fi # Fin # === set +o xtrace if [[ -n "$LOGFILE" ]]; then exec 1>&3 # Force all output to stdout and logs now exec 1> >( tee -a "${LOGFILE}" ) 2>&1 else # Force all output to stdout now exec 1>&3 fi # Using the cloud # --------------- echo "" echo "" echo "" # If you installed Horizon on this server you should be able # to access the site using your browser. if is_service_enabled horizon; then echo "Horizon is now available at http://$SERVICE_HOST/" fi # If Keystone is present you can point ``nova`` cli to this server if is_service_enabled key; then echo "Keystone is serving at $KEYSTONE_AUTH_PROTOCOL://$SERVICE_HOST:$KEYSTONE_API_PORT/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 # Echo ``HOST_IP`` - useful for ``build_uec.sh``, which uses dhcp to give the instance an address echo "This is your host ip: $HOST_IP" # Warn that ``EXTRA_FLAGS`` needs to be converted to ``EXTRA_OPTS`` if [[ -n "$EXTRA_FLAGS" ]]; then echo_summary "WARNING: EXTRA_FLAGS is defined and may need to be converted to EXTRA_OPTS" fi # Indicate how long this took to run (bash maintained variable ``SECONDS``) echo_summary "stack.sh completed in $SECONDS seconds."