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