training-labs/labs/osbash/scripts/test/launch_instance.sh

#!/usr/bin/env bash
set -o errexit -o nounset
TOP_DIR=$(cd "$(dirname "$0")/.." && pwd)
source "$TOP_DIR/config/paths"
source "$CONFIG_DIR/credentials"
source "$LIB_DIR/functions.guest.sh"
source "$CONFIG_DIR/demo-openstackrc.sh"

exec_logfile

indicate_current_auto

#------------------------------------------------------------------------------
# Launch a demo instance.
#------------------------------------------------------------------------------

# Packets from the instance VM destined for the Internet will have its
# floating IP address as the sender address. For your instance VM to
# get Internet access, you will probably have to configure masquerading
# on your host computer.

# On Linux, turning on masquerading may look something like this:

# echo "1" > /proc/sys/net/ipv4/ip_forward
# modprobe ip_tables
# modprobe ip_conntrack
# iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE
# iptables -A FORWARD -i eth0 -o vboxnet2 -m state \
#          --state RELATED,ESTABLISHED -j ACCEPT
# iptables -A FORWARD -i vboxnet2 -o eth0 -j ACCEPT

# Set this true if you have masquerading enabled to allow instance VMs access
# to the Internet.
: ${MASQUERADING:=true}

# Set this true if you want the instance to use the Google Public DNS name
# server. The default uses dnsmasq running on a node.
: ${EXT_DNS:=true}

DEMO_INSTANCE_NAME=demo-instance1

echo "SUM --- BEGIN"

function ssh_no_chk_node {
    ssh_no_chk -i "$HOME/.ssh/osbash_key" "$@"
}

function ssh_no_chk {
    echo "ssh $@"
    # Options set to disable strict host key checking and related messages.
    ssh \
        -o "UserKnownHostsFile /dev/null" \
        -o "StrictHostKeyChecking no" \
        -o LogLevel=error \
        "$@"
}

# Work around neutron client failing with unsupported locale settings
if [[ "$(neutron --help)" == "unsupported locale setting" ]]; then
    echo "Locale not supported on node, setting LC_ALL=C."
    export LC_ALL=C
fi

function wait_for_service {
    local node=$1
    local service=$2
    local cnt=0
    echo -n "Node $node, service $service:"
    until ssh_no_chk_node "$node" service "$service" status | \
            grep -q "start/running"; do
        cnt=$((cnt + 1))
        if [ $((cnt % 150)) -eq 0 ]; then
            echo " does not seem to come up. Forcing restart."

            echo
            echo "SUM ERROR $service on node $node not coming up."
            ssh_no_chk_node "$node" \
                sudo service "$service" restart
            SERVICE_RESTARTS="${SERVICE_RESTARTS:-""}$service@$node "
        fi
        sleep 2
        echo -n .
    done
    echo " up"
}

echo "Running on host: $(hostname)"

echo "Checking network connection to network node."
ping -c1 network-mgmt
echo

echo "Checking network connection to compute node."
ping -c1 compute-mgmt
echo

echo "Checking services on network node."
wait_for_service network-mgmt openvswitch-switch
wait_for_service network-mgmt neutron-plugin-openvswitch-agent
wait_for_service network-mgmt neutron-l3-agent
wait_for_service network-mgmt neutron-dhcp-agent
wait_for_service network-mgmt neutron-metadata-agent
echo

echo "Checking services on compute node."
wait_for_service compute-mgmt nova-compute
wait_for_service compute-mgmt openvswitch-switch
wait_for_service compute-mgmt neutron-plugin-openvswitch-agent
echo

function wait_for_nova_compute {
    if sudo nova-manage service list --service nova-compute | \
            grep -q ":-)"; then
        return 0
    fi
    echo "  Waiting for nova-compute to switch from XXX to :-)."
    if ssh_no_chk_node compute-mgmt service nova-compute status | \
            grep -q "start/running"; then
        echo -n "  Service is up, waiting (may take a few minutes)."
    fi
    local cnt=0
    local start=$(date +%s)
    while sudo nova-manage service list --service nova-compute | grep -q XXX; do
        cnt=$((cnt + 1))
        sleep 5
        if ssh_no_chk_node compute-mgmt service nova-compute status | \
                grep -q "start/running"; then
            if [ $cnt -eq 300 ]; then
                # This should never happen.
                echo "SUM ABORT nova-compute remains XXX while up."
                echo "Aborting."
                exit 1
            fi
            echo -n .
        else
            echo
            echo "SUM ERROR nova-compute on compute node has died."
            echo "Restarting nova-compute on compute node."
            ssh_no_chk_node compute-mgmt \
                sudo service nova-compute restart
            NOVA_COMPUTE_RESTART=$((${NOVA_COMPUTE_RESTART:-0} + 1))
        fi
    done
    echo
}

function wait_for_nova_services {
    local start=$(date +%s)

    echo "Checking services in sudo nova-manage service list."
    echo -n "  Waiting for controller services to switch from XXX to :-)."
    # Ignore nova-compute for now, even if a custom config has it on controller
    while sudo nova-manage service list --host controller | \
        grep -v nova-compute | grep -q XXX; do
        sleep 2
        echo -n .
    done
    echo

    if ! sudo nova-manage service list | grep -q nova-compute; then
        echo -n "  Waiting for nova-compute to turn up in list."
        until sudo nova-manage service list | grep -q nova-compute; do
            sleep 2
            echo -n .
        done
        echo
    fi

    wait_for_nova_compute
    echo
    echo "SUM wait for nova services: $(($(date +%s) - start))"
}

if [ ${NOVA_COMPUTE_RESTART:-0} -ne 0 ]; then
    echo "SUM ERROR nova-compute restarts: $NOVA_COMPUTE_RESTART"
fi

wait_for_nova_services

echo "All services are ready:"
sudo nova-manage service list
echo

function show_compute_resource_usage {
    echo "nova list:"
    nova list
    (
    source "$CONFIG_DIR/admin-openstackrc.sh"
    echo "As admin user, nova host-list:"
    nova host-list
    echo "As admin user, nova host-describe compute:"
    nova host-describe compute
    )
}

function wait_for_neutron_agents {
    local agent_list=$LOG_DIR/test-agent.list
    local start=$(date +%s)
    echo -n "Waiting for agents in neutron agent-list."
    (
    source "$CONFIG_DIR/admin-openstackrc.sh"
    neutron agent-list | sort > "$agent_list"
    local out=$(grep " :-)  " "$agent_list" || rc=$?)
    if [ -n "$out" ]; then
        echo
        echo "$out"
    fi
    while [ : ]; do
        neutron agent-list | sort > "$agent_list.new"
        out=$(comm -13 "$agent_list" "$agent_list.new")
        if [ -n "$out" ]; then
            echo
            echo "$out"
        fi
        if ! grep -q " xxx  " "$agent_list"; then
            break
        fi
        mv "$agent_list.new" "$agent_list"
        sleep 1
        echo -n .
    done
    echo
    echo "All agents are ready."
    neutron agent-list
    echo
    )
    echo "SUM wait for neutron agents: $(($(date +%s) - start))"
}

wait_for_neutron_agents

function check_namespaces {
    local cnt

    echo -n "Getting router namespace."
    cnt=0
    until ssh_no_chk_node network-mgmt ip netns | grep qrouter; do
        cnt=$((cnt + 1))
        sleep 1
        echo -n "."
    done
    echo "SUM wait for router namespace: $cnt"
    local nsrouter=$(ssh_no_chk_node network-mgmt ip netns | grep qrouter)

    echo -n "Getting DHCP namespace."
    cnt=0
    until ssh_no_chk_node network-mgmt ip netns | grep qdhcp; do
        cnt=$((cnt + 1))
        if [ $cnt -eq 10 ]; then
            echo
            echo "SUM ERROR No DCHP namespace, restarting neutron-dhcp-agent."
            echo "Restarting neutron-dhcp-agent on network node."
            ssh_no_chk_node network-mgmt \
                sudo service neutron-dhcp-agent restart
        fi
        sleep 1
        echo -n "."
    done
    echo "SUM wait for DHCP namespace: $cnt"
    local nsdhcp=$(ssh_no_chk_node network-mgmt ip netns | grep qdhcp)

    echo -n "Waiting for interface qr-* in router namespace."
    cnt=0
    until ssh_no_chk_node network-mgmt \
            sudo ip netns exec "$nsrouter" ip addr | \
            grep -Po "(?<=: )qr-.*(?=:)"; do
        cnt=$((cnt + 1))
        sleep 1
        echo -n "."
    done
    echo "SUM wait for interface qr-*: $cnt"

    echo -n "Waiting for interface qg-* in router namespace."
    cnt=0
    until ssh_no_chk_node network-mgmt \
            sudo ip netns exec "$nsrouter" ip addr | \
            grep -Po "(?<=: )qg-.*(?=:)"; do
        cnt=$((cnt + 1))
        sleep 1
        echo -n "."
    done
    echo "SUM wait for interface qg-*: $cnt"

    echo -n "Waiting for interface tap* in DHCP namespace."
    cnt=0
    until ssh_no_chk_node network-mgmt \
            sudo ip netns exec "$nsdhcp" ip addr | \
            grep -Po "(?<=: )tap.*(?=:)"; do
        cnt=$((cnt + 1))
        sleep 1
        echo -n "."
    done
    echo "SUM wait for interface tap*: $cnt"
}

check_namespaces

if [ ! -f ~/.ssh/id_rsa ]; then
    echo "Generating an ssh key pair (saved to ~/.ssh/id_rsa*)."
    # For training cluster: no password protection on keys to make scripting
    # easier
    ssh-keygen -f ~/.ssh/id_rsa -N ""
fi

function check_demo_key {
    echo -n "Checking if 'demo-key' is already in our OpenStack environment: "
    if nova keypair-show demo-key >/dev/null 2>&1; then
        echo "yes."

        echo -n "Checking if the 'demo-key' key pair matches our ssh key: "

        ssh_key=$(< ~/.ssh/id_rsa.pub awk '{print $2}')
        stored_key=$(nova keypair-show demo-key | \
            awk '/^Public key: ssh-rsa/ {print $4}')

        if [ "$ssh_key" != "$stored_key" ]; then
            echo "no."
            echo "Removing the 'demo-key' from the OpenStack envirnoment."
            nova keypair-delete demo-key
        else
            echo "yes."
        fi
    else
        echo "no."
    fi
}
check_demo_key

if ! nova keypair-show demo-key 2>/dev/null; then
    echo "Adding the public key to our OpenStack environment."
    nova keypair-add --pub-key ~/.ssh/id_rsa.pub demo-key
fi

echo "Verifying addition of the public key."
nova keypair-list

echo "Listing available flavors."
nova flavor-list

echo "Listing available images."
nova image-list

# Wait for neutron to start
wait_for_neutron

echo "Listing available networks."
neutron net-list

DEMO_NET_ID=$(neutron net-list | awk '/ demo-net / {print $2}')
echo "ID for demo-net tenant network: $DEMO_NET_ID"

echo "Listing available security groups."
nova secgroup-list

echo "Settings for demo-subnet:"
neutron subnet-show demo-subnet
echo

if [ "$EXT_DNS" = true ]; then
    echo "Setting DNS name server for subnet (passed to booting instance VMs)."
    neutron subnet-update demo-subnet --dns_nameservers list=true 8.8.4.4
    echo
else
    echo "Clearing DNS name server for subnet (passed to booting instance VMs)."
    neutron subnet-update demo-subnet --dns_nameservers action=clear
fi

echo "Settings for demo-subnet:"
neutron subnet-show demo-subnet
echo

nova list
nova list | awk " / $DEMO_INSTANCE_NAME / {print \$2}" | while read instance; do
    echo "Removing instance $DEMO_INSTANCE_NAME ($instance)."
    nova delete "$instance"
done
echo -n "Waiting for removed instances to disappear (may take > 1 min)."
while nova list|grep -q "$DEMO_INSTANCE_NAME"; do
    sleep 1
    echo -n .
done
echo

echo "There should be no $DEMO_INSTANCE_NAME instances left:"
nova list

NOVA_SCHED_LOG=/var/log/upstart/nova-scheduler.log
NOVA_API_LOG=/var/log/upstart/nova-api.log


VM_LAUNCHES=0

function request_instance {
    # Keep a copy of current state of nova-scheduler.log
    sudo cp -vf $NOVA_SCHED_LOG $NOVA_API_LOG /tmp

    if [ -n "${instance_info:-""}" ]; then
        rm -f "$instance_info"
    else
        instance_info=$LOG_DIR/test-instance.info
        echo "Instance info: $instance_info"
    fi

    local img_name=$(basename "$CIRROS_URL" -disk.img)

    echo "Requesting an instance."
    nova boot \
        --flavor m1.tiny \
        --image "$img_name" \
        --nic net-id="$DEMO_NET_ID" \
        --security-group default \
        --key-name demo-key \
        "$DEMO_INSTANCE_NAME" > "$instance_info"
    VM_LAUNCHES=$(( VM_LAUNCHES + 1 ))
}

BOOT_LOG=$LOG_DIR/test-instance.boot
echo "Boot log: $BOOT_LOG"

function save_boot_log {
    local rc=0
    rm -f "$BOOT_LOG"
    nova console-log "$DEMO_INSTANCE_NAME" >"$BOOT_LOG" 2>&1 || rc=$?
    if [ $rc -ne 0 ]; then
        echo >&2 "nova console-log returned error status $rc"
    fi
    return $rc
}

function explain_instance_failure {
    cat << TXT_INSTANCE_FAILURE

  After deleting an instance, it can take nova up to a minute to realize that
  the compute node is free. Under tight space constraints, this becomes a
  common source of failure.

  As an admin, we could list hosts (including compute hosts):

  $ nova host-list

  And check resource usage in description of host 'compute':

  $ nova host-describe compute

  As a regular user, we would have to keep trying for up to a minute and hope
  it works soon.

  The fastest way to update the database, however, is to restart nova-compute
  on the compute node.

TXT_INSTANCE_FAILURE
}

function status_409_fixed {
    echo "Checking log files for cause of failure."

    if sudo comm -13 /tmp/nova-scheduler.log $NOVA_SCHED_LOG |
            grep "has not been heard from in a while"; then
        echo
        echo "SUM ERROR Missing connection with nova-compute on compute node."
        echo "(Did controller node boot after compute node?)"
        echo
    elif sudo comm -13 /tmp/nova-scheduler.log $NOVA_SCHED_LOG |
            grep "Filter RamFilter returned 0 hosts"; then
        echo "SUM ERROR Filter RamFilter returned 0 hosts"
        explain_instance_failure
        show_compute_resource_usage
    elif sudo comm -13 /tmp/nova-api.log $NOVA_API_LOG |
            grep "HTTP exception thrown:"; then
        # Just waiting should be enough to fix this
        echo -n "Waiting for HTTP status 409 to cure itself."
        local cnt=0
        until [ $cnt -eq 5 ]; do
            if ! console_status_409; then
                HTTP_EXCEPTIONS="${HTTP_EXCEPTIONS:-""}$cnt "
                echo "okay"
                # We can continue with this instance
                return 0
            fi
            cnt=$((cnt + 1))
            sleep 2
            echo -n .
        done
        HTTP_EXCEPTIONS="${HTTP_EXCEPTIONS:-""}${cnt}-fail "
        echo "failed"
    else
        echo "Unknown reason. See for yourself."
        echo "nova-scheduler.log:"
        sudo comm -13 /tmp/nova-scheduler.log $NOVA_SCHED_LOG
        echo "nova-api.log:"
        sudo comm -13 /tmp/nova-api.log $NOVA_API_LOG
        echo "SUM ABORT Unknown 409 error"
        exit 1
    fi
    # Not fixed, need to try with new VM
    return 1
}

function console_status_409 {
    ! save_boot_log 2>/dev/null &&
        grep -q "is not ready (HTTP 409)" "$BOOT_LOG"
}

function console_status_404 {
    ! save_boot_log 2>/dev/null &&
        grep -q "Unable to get console (HTTP 404)" "$BOOT_LOG"
}

function instance_status {
    nova list | awk "/$DEMO_INSTANCE_NAME/ {print \$6}"
}

function instance_status_is {
    local status=$1
    nova list | grep "$DEMO_INSTANCE_NAME" | grep -q "$status"
}

while [ : ]; do
    echo "###################################################################"
    echo "Launching an instance VM ($VM_LAUNCHES)."
    request_instance > /dev/null

    if console_status_409; then
        echo "nova console-log returned:"
        cat "$BOOT_LOG"
        echo

        if ! status_409_fixed; then

            echo "Instance build failed."
            echo "Deleting failed instance VM."
            nova delete "$DEMO_INSTANCE_NAME"

            echo "Checking nova-compute on the compute node."
            wait_for_nova_compute

            echo -n "Requesting new instance VMs until it works."
            cnt=0
            while [ : ]; do
                request_instance >/dev/null
                if console_status_409; then
                    nova delete "$DEMO_INSTANCE_NAME"
                    cnt=$((cnt + 1))
                    if [ $cnt -eq 5 ]; then
                        echo
                        echo "SUM ERROR console status remains 409."
                        echo "Restarting nova-compute on compute node."
                        ssh_no_chk_node compute-mgmt \
                            sudo service nova-compute restart
                        NOVA_COMPUTE_RESTART=$((${NOVA_COMPUTE_RESTART:-0} + 1))
                    fi
                    sleep 2
                    echo -n .
                else
                    # Either no error or a different error
                    echo
                    break
                fi
            done
        fi
    fi

    if console_status_404; then
        echo "nova console-log returned:"
        cat "$BOOT_LOG"
        echo

        echo -n "Waiting for console."
        # Console status 404 may persist after instance status becomes ERROR.
        while console_status_404 && instance_status_is BUILD; do
            sleep 1
            echo -n .
        done
        echo
        if ! console_status_404; then
            echo "Console status is no longer 404."
        fi

    fi

    echo -n "Waiting for instance to get out of BUILD status."
    while instance_status_is BUILD; do
        sleep 1
        echo -n .
    done
    echo

    if instance_status_is ERROR; then
        echo "Instance VM status: ERROR"

        if sudo comm -13 /tmp/nova-scheduler.log $NOVA_SCHED_LOG |
                grep "Filter RetryFilter returned 0 hosts"; then
            echo "SUM ERROR RetryFilter returned 0 hosts"
            show_compute_resource_usage
            echo "Restarting nova-compute on compute node."
            ssh_no_chk_node compute-mgmt \
                sudo service nova-compute restart
            NOVA_COMPUTE_RESTART=$((${NOVA_COMPUTE_RESTART:-0} + 1))
        fi

        echo "Deleting failed instance VM."
        nova delete "$DEMO_INSTANCE_NAME"
    elif instance_status_is ACTIVE; then
        echo "Instance VM status: ACTIVE."
        break
    fi

    if [ $VM_LAUNCHES -eq 10 ]; then
        echo "SUM ABORT $VM_LAUNCHES launch attempts failed. Giving up."
        exit 1
    fi
done

if [ "${HTTP_EXCEPTIONS:-0}" != "0" ]; then
    echo "SUM ERROR HTTP exceptions: ${HTTP_EXCEPTIONS:-0}"
fi

echo -n "Waiting for DHCP discover."
until grep -q "Sending discover..." "$BOOT_LOG"; do
    sleep 2
    echo -n .
    save_boot_log
done
echo

echo -n "Waiting for DHCP success."
until grep -q "^Lease of" "$BOOT_LOG"; do
    DHCP_WAIT=$((${DHCP_WAIT:-0} + 1))
    if grep "No lease, failing" "$BOOT_LOG"; then
        echo "SUM ABORT DHCP wait: fail (${DHCP_WAIT:-0})"
        echo "Aborting."
        exit 1
    fi
    sleep 2
    echo -n .
    save_boot_log
done
echo
echo "SUM DHCP wait: ${DHCP_WAIT:-0}"
echo

echo -n "Waiting for metadata success."
until grep -q "successful after" "$BOOT_LOG"; do
    if grep "failed to read iid from metadata" "$BOOT_LOG"; then
        echo "SUM ABORT failed to get metadata"
        echo "Aborting."
        exit 1
    fi
    sleep 2
    echo -n .
    save_boot_log
done
echo

echo -n "Waiting for login prompt."
until grep -q "$DEMO_INSTANCE_NAME login:" "$BOOT_LOG"; do
    sleep 2
    echo -n .
    save_boot_log
done
echo

echo "Obtaining a VNC session URL for our instance."
nova get-vnc-console "$DEMO_INSTANCE_NAME" novnc

echo
echo "Permitting ICMP (ping) to our instances."
nova secgroup-add-rule default icmp -1 -1 0.0.0.0/0 2>/dev/null || rc=$?
if [ ${rc:-0} -ne 0 ]; then
    echo "Rule was already there."
fi

echo
echo "Permitting secure shell (SSH) access to our instances."
nova secgroup-add-rule default tcp 22 22 0.0.0.0/0 2>/dev/null || rc=$?
if [ ${rc:-0} -ne 0 ]; then
    echo "Rule was already there."
fi

echo
echo "Verifying security-group rules."
nova secgroup-list-rules default

echo
echo "Creating a floating IP address on the ext-net external network."
floating_ip_id=$(neutron floatingip-create ext-net | awk '/ id / {print $4}')
neutron floatingip-show "$floating_ip_id"

floating_ip=$(neutron floatingip-show "$floating_ip_id" |
    awk '/ floating_ip_address / {print $4}')

echo
echo "Associating the floating IP address with our instance."
nova floating-ip-associate "$DEMO_INSTANCE_NAME" "$floating_ip"

echo
echo "Checking the status of your floating IP address."
nova list

echo
echo -n "Verifying network connectivity to instance VM (may take 2+ min)."
# Since Juno, the floating IP often takes a long time to become pingable.
# Hopefully, this will be fixed, but for the time being we just ping the
# floating IP until we get a reply (or we reach a time limit and give up).
function patient_ping {
    local ip=$1
    local cnt=0

    while [ : ]; do
        echo -n .
        sleep 1

        # Ping the instance VM every ten seconds
        if [[ $((cnt % 10)) -eq 0 ]]; then
            if ping -c1 "$ip" > /dev/null ; then
                echo
                ping -c1 "$ip"
                echo "SUM ping instance VM after $cnt seconds."
                break
            fi
        fi

        # Abort if it takes too long
        if [[ $cnt -gt 600 ]]; then
            echo
            echo "SUM ERROR no ping for instance VM in $cnt seconds. Aborting."
            exit 1
        fi

        cnt=$((cnt + 1))
    done
}

patient_ping "$floating_ip"

echo
echo "Accessing our instance using SSH from the controller node."
ssh_no_chk "cirros@$floating_ip" uptime

echo
echo "Interface configuration on instance VM."
ssh_no_chk "cirros@$floating_ip" ip addr

echo
echo "Routing information on instance VM."
ssh_no_chk "cirros@$floating_ip" /sbin/route -n

echo
echo "/etc/resolv.conf on instance VM."
ssh_no_chk "cirros@$floating_ip" cat /etc/resolv.conf

echo
echo "Pinging our own floating IP from inside the instance."
ssh_no_chk "cirros@$floating_ip" ping -c1 "$floating_ip"

echo
echo "Pinging IP address of controller-api."
ssh_no_chk "cirros@$floating_ip" ping -c1 "$(hostname_to_ip controller-api)"

if [ "$EXT_DNS" = true ]; then
    echo "Skipping tests of dnsmasq /etc/hosts."
else
    # Works only with dnsmasq using the node's /etc/hosts
    echo
    echo "Pinging controller-api (test local DNS name resolution)."
    ssh_no_chk "cirros@$floating_ip" ping -c1 controller-api
    echo
    echo "Pinging network-api."
    ssh_no_chk "cirros@$floating_ip" ping -c1 network-api
fi

if [ "$MASQUERADING" = true -a "$EXT_DNS" = false ]; then
    echo
    echo "This may work thanks to masquerading."
    ssh_no_chk "cirros@$floating_ip" ping -c1 network-mgmt
    echo
    ssh_no_chk "cirros@$floating_ip" ping -c1 network-data
fi

function test_internet {
    if [ "$MASQUERADING" = true ]; then
        local ext_ping=1
        echo
        echo "Pinging Google Public DNS name server."
        until ssh_no_chk "cirros@$floating_ip" ping -c1 8.8.8.8; do
            if [ $ext_ping -eq 3 ]; then
                echo "Failed. Giving up."
                echo "SUM ERROR ping Internet: failed ($ext_ping)"
                ext_ping="$ext_ping (failed)"
                return 0
            fi
            echo
            echo "Trying again in 1 s."
            sleep 1
            ext_ping=$((ext_ping + 1))
        done
        echo "SUM ping Internet: $ext_ping"

        echo
        echo "Testing DNS name resolution within instance VM."
        ssh_no_chk "cirros@$floating_ip" ping -c1 openstack.org
    fi
}

test_internet

if [ "$EXT_DNS" = true ]; then
    echo
    echo "Removing DNS name servers from subnet."
    neutron subnet-update demo-subnet --dns_nameservers action=clear
fi

echo
echo "Summary"
echo "======="
echo "SUM service restarts: ${SERVICE_RESTARTS:--}"
echo "SUM instance launches: $VM_LAUNCHES"
echo "SUM END"

echo
echo "Try this, it should work:"
echo "Command: 'ssh cirros@$floating_ip' [ password: 'cubswin:)' ]"