tests/run.sh

   1 #!/bin/bash -e
   2 ##############################################################################
   3 # Copyright (c) 2016 NEC Corporation and others.
   4 #
   5 # All rights reserved. This program and the accompanying materials
   6 # are made available under the terms of the Apache License, Version 2.0
   7 # which accompanies this distribution, and is available at
   8 # http://www.apache.org/licenses/LICENSE-2.0
   9 ##############################################################################
  10
  11 # Configuration
  12
  13 [[ "${CI_DEBUG:-true}" == [Tt]rue ]] && set -x
  14
  15 IMAGE_URL=https://launchpad.net/cirros/trunk/0.3.0/+download/cirros-0.3.0-x86_64-disk.img
  16 #if an existing image name is provided in the enviroment, use that one
  17 IMAGE_NAME=${IMAGE_NAME:-cirros}
  18 IMAGE_FILE="${IMAGE_NAME}.img"
  19 IMAGE_FORMAT=qcow2
  20 VM_NAME=doctor_vm1
  21 VM_FLAVOR=m1.tiny
  22 ALARM_NAME=doctor_alarm1
  23 INSPECTOR_PORT=12345
  24 CONSUMER_PORT=12346
  25 DOCTOR_USER=doctor
  26 DOCTOR_PW=doctor
  27 DOCTOR_PROJECT=doctor
  28 #TODO: change back to `_member_` when JIRA DOCTOR-55 is done
  29 DOCTOR_ROLE=admin
  30
  31 TOP_DIR=$(cd $(dirname "$0") && pwd)
  32
  33 ssh_opts="-o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no"
  34 as_doctor_user="--os-username $DOCTOR_USER --os-password $DOCTOR_PW
  35                 --os-tenant-name $DOCTOR_PROJECT"
  36
  37
  38 # Functions
  39
  40 get_compute_host_info() {
  41     # get computer host info which VM boot in
  42     COMPUTE_HOST=$(openstack $as_doctor_user server show $VM_NAME |
  43                    grep "OS-EXT-SRV-ATTR:host" | awk '{ print $4 }')
  44     compute_host_in_undercloud=${COMPUTE_HOST%%.*}
  45     die_if_not_set $LINENO COMPUTE_HOST "Failed to get compute hostname"
  46
  47     if is_installer apex; then
  48         COMPUTE_USER=${COMPUTE_USER:-heat-admin}
  49         COMPUTE_IP=$(sudo ssh $ssh_opts $INSTALLER_IP \
  50              "source stackrc; \
  51              nova show $compute_host_in_undercloud \
  52              | awk '/ ctlplane network /{print \$5}'")
  53     elif is_installer fuel; then
  54         COMPUTE_USER=${COMPUTE_USER:-root}
  55         node_id=$(echo $compute_host_in_undercloud | cut -d "-" -f 2)
  56         COMPUTE_IP=$(sshpass -p r00tme ssh 2>/dev/null $ssh_opts root@${INSTALLER_IP} \
  57              "fuel node|awk -F '|' -v id=$node_id '{if (\$1 == id) print \$5}' |xargs")
  58     elif is_installer local; then
  59         COMPUTE_USER=${COMPUTE_USER:-$(whoami)}
  60         COMPUTE_IP=$(getent hosts "$COMPUTE_HOST" | awk '{ print $1 }')
  61     fi
  62
  63     die_if_not_set $LINENO COMPUTE_IP "Could not resolve $COMPUTE_HOST. Either manually set COMPUTE_IP or enable DNS resolution."
  64     echo "COMPUTE_HOST=$COMPUTE_HOST"
  65     echo "COMPUTE_IP=$COMPUTE_IP"
  66
  67     # verify connectivity to target compute host
  68     ping -c 1 "$COMPUTE_IP"
  69     if [[ $? -ne 0 ]] ; then
  70         die $LINENO "Can not ping to computer host"
  71     fi
  72
  73     # verify ssh to target compute host
  74     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'exit'
  75     if [[ $? -ne 0 ]] ; then
  76         die $LINENO "Can not ssh to computer host"
  77     fi
  78 }
  79
  80 get_consumer_ip() {
  81     local get_consumer_command="ip route get $COMPUTE_IP | awk '/ src /{print \$NF}'"
  82     if is_installer apex; then
  83         CONSUMER_IP=$(sudo ssh $ssh_opts root@$INSTALLER_IP \
  84                       "$get_consumer_command")
  85     elif is_installer fuel; then
  86         CONSUMER_IP=$(sudo sshpass -p r00tme ssh $ssh_opts root@${INSTALLER_IP} \
  87                       "$get_consumer_command")
  88     elif is_installer local; then
  89         CONSUMER_IP=`$get_consumer_command`
  90     fi
  91     echo "CONSUMER_IP=$CONSUMER_IP"
  92
  93     die_if_not_set $LINENO CONSUMER_IP "Could not get CONSUMER_IP."
  94 }
  95
  96 download_image() {
  97     #if a different name was provided for the image in the enviroment there's no need to download the image
  98     use_existing_image=false
  99     openstack image list | grep -q " $IMAGE_NAME " && use_existing_image=true
 100
 101     if [[ "$use_existing_image" == false ]] ; then
 102         [ -e "$IMAGE_FILE" ] && return 0
 103         wget "$IMAGE_URL" -o "$IMAGE_FILE"
 104     fi
 105 }
 106
 107 register_image() {
 108     openstack image list | grep -q " $IMAGE_NAME " && return 0
 109     openstack image create "$IMAGE_NAME" \
 110                            --public \
 111                            --disk-format "$IMAGE_FORMAT" \
 112                            --container-format bare \
 113                            --file "$IMAGE_FILE"
 114 }
 115
 116 create_test_user() {
 117     openstack project list | grep -q " $DOCTOR_PROJECT " || {
 118         openstack project create "$DOCTOR_PROJECT"
 119     }
 120     openstack user list | grep -q " $DOCTOR_USER " || {
 121         openstack user create "$DOCTOR_USER" --password "$DOCTOR_PW" \
 122                               --project "$DOCTOR_PROJECT"
 123     }
 124     openstack user role list "$DOCTOR_USER" --project "$DOCTOR_PROJECT" \
 125     | grep -q " $DOCTOR_ROLE " || {
 126         openstack role add "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 127                            --project "$DOCTOR_PROJECT"
 128     }
 129 }
 130
 131 boot_vm() {
 132     # test VM done with test user, so can test non-admin
 133     openstack $as_doctor_user server list | grep -q " $VM_NAME " && return 0
 134     openstack $as_doctor_user server create --flavor "$VM_FLAVOR" \
 135                             --image "$IMAGE_NAME" \
 136                             "$VM_NAME"
 137     sleep 1
 138 }
 139
 140 create_alarm() {
 141     # get vm_id as test user
 142     ceilometer $as_doctor_user alarm-list | grep -q " $ALARM_NAME " && return 0
 143     vm_id=$(openstack $as_doctor_user server list | grep " $VM_NAME " | awk '{print $2}')
 144     # TODO(r-mibu): change notification endpoint from localhost to the consumer
 145     # IP address (functest container).
 146     ceilometer $as_doctor_user alarm-event-create --name "$ALARM_NAME" \
 147         --alarm-action "http://localhost:$CONSUMER_PORT/failure" \
 148         --description "VM failure" \
 149         --enabled True \
 150         --repeat-actions False \
 151         --severity "moderate" \
 152         --event-type compute.instance.update \
 153         -q "traits.state=string::error; traits.instance_id=string::$vm_id"
 154 }
 155
 156 start_monitor() {
 157     pgrep -f "python monitor.py" && return 0
 158     sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
 159         "http://127.0.0.1:$INSPECTOR_PORT/events" > monitor.log 2>&1 &
 160 }
 161
 162 stop_monitor() {
 163     pgrep -f "python monitor.py" || return 0
 164     sudo kill $(pgrep -f "python monitor.py")
 165 }
 166
 167 start_consumer() {
 168     pgrep -f "python consumer.py" && return 0
 169     python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 &
 170
 171     # NOTE(r-mibu): create tunnel to the controller nodes, so that we can
 172     # avoid some network problems dpends on infra and installers.
 173     # This tunnel will be terminated by stop_consumer() or after 10 mins passed.
 174     if ! is_installer local; then
 175         if is_installer apex; then
 176             CONTROLLER_IPS=$(sudo ssh $ssh_opts $INSTALLER_IP \
 177                              "source stackrc; \
 178                              nova list | grep ' overcloud-controller-[0-9] ' \
 179                              | sed -e 's/^.*ctlplane=//' -e 's/ *|\$//'")
 180         elif is_installer fuel; then
 181             CONTROLLER_IPS=$(sshpass -p r00tme ssh 2>/dev/null $ssh_opts root@${INSTALLER_IP} \
 182                             "fuel node | grep controller | cut -d '|' -f 5|xargs")
 183         fi
 184
 185         die_if_not_set $LINENO CONTROLLER_IPS "Could not get CONTROLLER_IPS."
 186         for ip in $CONTROLLER_IPS
 187         do
 188             forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
 189             tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
 190             $tunnel_command > "ssh_tunnel.${ip}.log" 2>&1 < /dev/null &
 191         done
 192     fi
 193 }
 194
 195 stop_consumer() {
 196     pgrep -f "python consumer.py" || return 0
 197     kill $(pgrep -f "python consumer.py")
 198
 199     # NOTE(r-mibu): terminate tunnels to the controller nodes
 200     if ! is_installer local; then
 201         for ip in $CONTROLLER_IPS
 202         do
 203             forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
 204             tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
 205             kill $(pgrep -f "$tunnel_command")
 206         done
 207     fi
 208 }
 209
 210 wait_for_vm_launch() {
 211     echo "waiting for vm launch..."
 212
 213     count=0
 214     while [[ ${count} -lt 60 ]]
 215     do
 216         state=$(openstack $as_doctor_user server list | grep " $VM_NAME " | awk '{print $6}')
 217         if [[ "$state" == "ACTIVE" ]]; then
 218             # NOTE(cgoncalves): sleeping for a bit to stabilize
 219             # See python-openstackclient/functional/tests/compute/v2/test_server.py:wait_for_status
 220             sleep 5
 221             return 0
 222         fi
 223         if [[ "$state" == "ERROR" ]]; then
 224             die $LINENO "vm state is ERROR"
 225         fi
 226         count=$(($count+1))
 227         sleep 1
 228     done
 229     die $LINENO "Time out while waiting for VM launch"
 230 }
 231
 232 inject_failure() {
 233     echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..."
 234     cat > disable_network.sh << 'END_TXT'
 235 #!/bin/bash -x
 236 dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $7}')
 237 sleep 1
 238 sudo ip link set $dev down
 239 echo "doctor set host down at" $(date "+%s.%N")
 240 sleep 180
 241 sudo ip link set $dev up
 242 sleep 1
 243 END_TXT
 244     sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh
 245     chmod +x disable_network.sh
 246     scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:"
 247     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &'
 248 }
 249
 250 profile_performance_poc() {
 251     total=`python -c "print(int(($notified-$detected)*1000))"`
 252
 253     export DOCTOR_PROFILER_T00=0
 254     export DOCTOR_PROFILER_T09=$((total))
 255     python profiler-poc.py
 256 }
 257
 258 calculate_notification_time() {
 259     detected=$(grep "doctor monitor detected at" monitor.log | awk '{print $10}')
 260     notified=$(grep "doctor consumer notified at" consumer.log | awk '{print $10}')
 261     if ! grep -q "doctor consumer notified at" consumer.log ; then
 262         die $LINENO "Consumer hasn't received fault notification."
 263     fi
 264
 265     if [[ PROFILER == 'poc' ]]; then
 266         profile_performance_poc
 267     fi
 268
 269     echo "$notified $detected" | \
 270         awk '{
 271             d = $1 - $2;
 272             if (d < 1 && d > 0) { print d " OK"; exit 0 }
 273             else { print d " NG"; exit 1 }
 274         }'
 275 }
 276
 277 check_host_status() {
 278     expected_state=$1
 279
 280     host_status_line=$(openstack $as_doctor_user --os-compute-api-version 2.16 \
 281                        server show $VM_NAME | grep "host_status")
 282     host_status=$(echo $host_status_line | awk '{print $4}')
 283     die_if_not_set $LINENO host_status "host_status not reported by: nova show $VM_NAME"
 284     if [[ "$expected_state" =~ "$host_status" ]] ; then
 285         echo "$VM_NAME showing host_status: $host_status"
 286     else
 287         die $LINENO "host_status:$host_status not equal to expected_state: $expected_state"
 288     fi
 289 }
 290
 291 cleanup() {
 292     set +e
 293     echo "cleanup..."
 294     stop_monitor
 295     stop_inspector
 296     stop_consumer
 297
 298     echo "waiting disabled compute host back to be enabled..."
 299     python ./nova_force_down.py "$COMPUTE_HOST" --unset
 300     sleep 240
 301     check_host_status "UP"
 302     scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
 303
 304     openstack $as_doctor_user server list | grep -q " $VM_NAME " && openstack $as_doctor_user server delete "$VM_NAME"
 305     sleep 1
 306     alarm_id=$(ceilometer $as_doctor_user alarm-list | grep " $ALARM_NAME " | awk '{print $2}')
 307     sleep 1
 308     [ -n "$alarm_id" ] && ceilometer $as_doctor_user alarm-delete "$alarm_id"
 309     sleep 1
 310
 311     image_id=$(openstack image list | grep " $IMAGE_NAME " | awk '{print $2}')
 312     sleep 1
 313     #if an existing image was used, there's no need to remove it here
 314     if [[ "$use_existing_image" == false ]] ; then
 315         [ -n "$image_id" ] && openstack image delete "$image_id"
 316     fi
 317     openstack role remove "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 318                               --project "$DOCTOR_PROJECT"
 319     openstack project delete "$DOCTOR_PROJECT"
 320     openstack user delete "$DOCTOR_USER"
 321
 322     cleanup_installer
 323     cleanup_inspector
 324 }
 325
 326 # Main process
 327
 328 echo "Note: doctor/tests/run.sh has been executed."
 329
 330 trap cleanup EXIT
 331
 332 source $TOP_DIR/functions-common
 333 source $TOP_DIR/lib/installer
 334 source $TOP_DIR/lib/inspector
 335
 336 setup_installer
 337
 338 echo "preparing VM image..."
 339 download_image
 340 register_image
 341
 342 echo "creating test user..."
 343 create_test_user
 344
 345 echo "creating VM..."
 346 boot_vm
 347 wait_for_vm_launch
 348
 349 echo "get computer host info..."
 350 get_compute_host_info
 351
 352 echo "creating alarm..."
 353 #TODO: change back to use, network problems depends on infra and installers
 354 #get_consumer_ip
 355 create_alarm
 356
 357 echo "starting doctor sample components..."
 358 start_inspector
 359 start_monitor
 360 start_consumer
 361
 362 sleep 60
 363 echo "injecting host failure..."
 364 inject_failure
 365 sleep 60
 366
 367 check_host_status "(DOWN|UNKNOWN)"
 368 calculate_notification_time
 369
 370 echo "done"