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="${IMAGE_NAME}.img"
  19 IMAGE_FORMAT=qcow2
  20 VM_NAME=doctor_vm1
  21 VM_FLAVOR=m1.tiny
  22 ALARM_NAME=doctor_alarm1
  23 INSPECTOR_PORT=12345
  24 CONSUMER_PORT=12346
  25 DOCTOR_USER=doctor
  26 DOCTOR_PW=doctor
  27 DOCTOR_PROJECT=doctor
  28 #TODO: change back to `_member_` when JIRA DOCTOR-55 is done
  29 DOCTOR_ROLE=admin
  30 PROFILER_TYPE=${PROFILER_TYPE:-none}
  31
  32 TOP_DIR=$(cd $(dirname "$0") && pwd)
  33
  34 ssh_opts="-o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no"
  35 as_doctor_user="--os-username $DOCTOR_USER --os-password $DOCTOR_PW
  36                 --os-tenant-name $DOCTOR_PROJECT"
  37
  38
  39 # Functions
  40
  41 get_compute_host_info() {
  42     # get computer host info which VM boot in
  43     COMPUTE_HOST=$(openstack $as_doctor_user server show $VM_NAME |
  44                    grep "OS-EXT-SRV-ATTR:host" | awk '{ print $4 }')
  45     compute_host_in_undercloud=${COMPUTE_HOST%%.*}
  46     die_if_not_set $LINENO COMPUTE_HOST "Failed to get compute hostname"
  47
  48     if is_installer apex; then
  49         COMPUTE_USER=${COMPUTE_USER:-heat-admin}
  50         COMPUTE_IP=$(sudo ssh $ssh_opts $INSTALLER_IP \
  51              "source stackrc; \
  52              nova show $compute_host_in_undercloud \
  53              | awk '/ ctlplane network /{print \$5}'")
  54     elif is_installer fuel; then
  55         COMPUTE_USER=${COMPUTE_USER:-root}
  56         node_id=$(echo $compute_host_in_undercloud | cut -d "-" -f 2)
  57         COMPUTE_IP=$(sshpass -p r00tme ssh 2>/dev/null $ssh_opts root@${INSTALLER_IP} \
  58              "fuel node|awk -F '|' -v id=$node_id '{if (\$1 == id) print \$5}' |xargs")
  59     elif is_installer local; then
  60         COMPUTE_USER=${COMPUTE_USER:-$(whoami)}
  61         COMPUTE_IP=$(getent hosts "$COMPUTE_HOST" | awk '{ print $1 }')
  62     fi
  63
  64     die_if_not_set $LINENO COMPUTE_IP "Could not resolve $COMPUTE_HOST. Either manually set COMPUTE_IP or enable DNS resolution."
  65     echo "COMPUTE_HOST=$COMPUTE_HOST"
  66     echo "COMPUTE_IP=$COMPUTE_IP"
  67
  68     # verify connectivity to target compute host
  69     ping -c 1 "$COMPUTE_IP"
  70     if [[ $? -ne 0 ]] ; then
  71         die $LINENO "Can not ping to computer host"
  72     fi
  73
  74     # verify ssh to target compute host
  75     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'exit'
  76     if [[ $? -ne 0 ]] ; then
  77         die $LINENO "Can not ssh to computer host"
  78     fi
  79 }
  80
  81 get_consumer_ip() {
  82     local get_consumer_command="ip route get $COMPUTE_IP | awk '/ src /{print \$NF}'"
  83     if is_installer apex; then
  84         CONSUMER_IP=$(sudo ssh $ssh_opts root@$INSTALLER_IP \
  85                       "$get_consumer_command")
  86     elif is_installer fuel; then
  87         CONSUMER_IP=$(sudo sshpass -p r00tme ssh $ssh_opts root@${INSTALLER_IP} \
  88                       "$get_consumer_command")
  89     elif is_installer local; then
  90         CONSUMER_IP=`$get_consumer_command`
  91     fi
  92     echo "CONSUMER_IP=$CONSUMER_IP"
  93
  94     die_if_not_set $LINENO CONSUMER_IP "Could not get CONSUMER_IP."
  95 }
  96
  97 download_image() {
  98     #if a different name was provided for the image in the enviroment there's no need to download the image
  99     use_existing_image=false
 100     openstack image list | grep -q " $IMAGE_NAME " && use_existing_image=true
 101
 102     if [[ "$use_existing_image" == false ]] ; then
 103         [ -e "$IMAGE_FILE" ] && return 0
 104         wget "$IMAGE_URL" -o "$IMAGE_FILE"
 105     fi
 106 }
 107
 108 register_image() {
 109     openstack image list | grep -q " $IMAGE_NAME " && return 0
 110     openstack image create "$IMAGE_NAME" \
 111                            --public \
 112                            --disk-format "$IMAGE_FORMAT" \
 113                            --container-format bare \
 114                            --file "$IMAGE_FILE"
 115 }
 116
 117 create_test_user() {
 118     openstack project list | grep -q " $DOCTOR_PROJECT " || {
 119         openstack project create "$DOCTOR_PROJECT"
 120     }
 121     openstack user list | grep -q " $DOCTOR_USER " || {
 122         openstack user create "$DOCTOR_USER" --password "$DOCTOR_PW" \
 123                               --project "$DOCTOR_PROJECT"
 124     }
 125     openstack user role list "$DOCTOR_USER" --project "$DOCTOR_PROJECT" \
 126     | grep -q " $DOCTOR_ROLE " || {
 127         openstack role add "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 128                            --project "$DOCTOR_PROJECT"
 129     }
 130 }
 131
 132 boot_vm() {
 133     # test VM done with test user, so can test non-admin
 134     openstack $as_doctor_user server list | grep -q " $VM_NAME " && return 0
 135     openstack $as_doctor_user server create --flavor "$VM_FLAVOR" \
 136                             --image "$IMAGE_NAME" \
 137                             "$VM_NAME"
 138     sleep 1
 139 }
 140
 141 create_alarm() {
 142     # get vm_id as test user
 143     ceilometer $as_doctor_user alarm-list | grep -q " $ALARM_NAME " && return 0
 144     vm_id=$(openstack $as_doctor_user server list | grep " $VM_NAME " | awk '{print $2}')
 145     # TODO(r-mibu): change notification endpoint from localhost to the consumer
 146     # IP address (functest container).
 147     ceilometer $as_doctor_user alarm-event-create --name "$ALARM_NAME" \
 148         --alarm-action "http://localhost:$CONSUMER_PORT/failure" \
 149         --description "VM failure" \
 150         --enabled True \
 151         --repeat-actions False \
 152         --severity "moderate" \
 153         --event-type compute.instance.update \
 154         -q "traits.state=string::error; traits.instance_id=string::$vm_id"
 155 }
 156
 157 start_monitor() {
 158     pgrep -f "python monitor.py" && return 0
 159     sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
 160         "http://127.0.0.1:$INSPECTOR_PORT/events" > monitor.log 2>&1 &
 161 }
 162
 163 stop_monitor() {
 164     pgrep -f "python monitor.py" || return 0
 165     sudo kill $(pgrep -f "python monitor.py")
 166 }
 167
 168 start_consumer() {
 169     pgrep -f "python consumer.py" && return 0
 170     python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 &
 171
 172     # NOTE(r-mibu): create tunnel to the controller nodes, so that we can
 173     # avoid some network problems dpends on infra and installers.
 174     # This tunnel will be terminated by stop_consumer() or after 10 mins passed.
 175     if ! is_installer local; then
 176         if is_installer apex; then
 177             CONTROLLER_IPS=$(sudo ssh $ssh_opts $INSTALLER_IP \
 178                              "source stackrc; \
 179                              nova list | grep ' overcloud-controller-[0-9] ' \
 180                              | sed -e 's/^.*ctlplane=//' -e 's/ *|\$//'")
 181         elif is_installer fuel; then
 182             get_controller_ips
 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             openstack $as_doctor_user server show $VM_NAME
 225             die $LINENO "vm state is ERROR"
 226         fi
 227         count=$(($count+1))
 228         sleep 1
 229     done
 230     die $LINENO "Time out while waiting for VM launch"
 231 }
 232
 233 inject_failure() {
 234     echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..."
 235     cat > disable_network.sh << 'END_TXT'
 236 #!/bin/bash -x
 237 dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $7}')
 238 [[ -n "$dev" ]] || dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $5}')
 239 sleep 1
 240 sudo ip link set $dev down
 241 echo "doctor set link down at" $(date "+%s.%N")
 242 sleep 180
 243 sudo ip link set $dev up
 244 sleep 1
 245 END_TXT
 246     sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh
 247     chmod +x disable_network.sh
 248     scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:"
 249     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &'
 250     # use host time to get rid of potential time sync deviation between nodes
 251     triggered=$(date "+%s.%N")
 252 }
 253
 254 calculate_notification_time() {
 255     if ! grep -q "doctor consumer notified at" consumer.log ; then
 256         die $LINENO "Consumer hasn't received fault notification."
 257     fi
 258
 259     #keep 'at' as the last keyword just before the value, and
 260     #use regex to get value instead of the fixed column
 261     detected=$(grep "doctor monitor detected at" monitor.log |\
 262                sed -e "s/^.* at //")
 263     notified=$(grep "doctor consumer notified at" consumer.log |\
 264                sed -e "s/^.* at //")
 265
 266     echo "$notified $detected" | \
 267         awk '{
 268             d = $1 - $2;
 269             if (d < 1 && d > 0) { print d " OK"; exit 0 }
 270             else { print d " NG"; exit 1 }
 271         }'
 272 }
 273
 274 check_host_status() {
 275     expected_state=$1
 276
 277     host_status_line=$(openstack $as_doctor_user --os-compute-api-version 2.16 \
 278                        server show $VM_NAME | grep "host_status")
 279     host_status=$(echo $host_status_line | awk '{print $4}')
 280     die_if_not_set $LINENO host_status "host_status not reported by: nova show $VM_NAME"
 281     if [[ "$expected_state" =~ "$host_status" ]] ; then
 282         echo "$VM_NAME showing host_status: $host_status"
 283     else
 284         die $LINENO "host_status:$host_status not equal to expected_state: $expected_state"
 285     fi
 286 }
 287
 288 unset_forced_down_hosts() {
 289     for host in $(openstack compute service list --service nova-compute \
 290                   -f value -c Host -c State | sed -n -e '/down$/s/ *down$//p')
 291     do
 292         # TODO (r-mibu): make sample inspector use keystone v3 api
 293         OS_AUTH_URL=${OS_AUTH_URL/v3/v2.0} \
 294         python ./nova_force_down.py $host --unset
 295     done
 296
 297     echo "waiting disabled compute host back to be enabled..."
 298     wait_until 'openstack compute service list --service nova-compute
 299                 -f value -c State | grep -q down' 240 5
 300 }
 301
 302 collect_logs() {
 303     unset_forced_down_hosts
 304     # TODO: We need to make sure the target compute host is back to IP
 305     #       reachable. wait_ping() will be added by tojuvone .
 306     sleep 110
 307     scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
 308
 309     # TODO(yujunz) collect other logs, e.g. nova, aodh
 310 }
 311
 312 run_profiler() {
 313     if [[ "$PROFILER_TYPE" == "poc" ]]; then
 314         linkdown=$(grep "doctor set link down at " disable_network.log |\
 315                   sed -e "s/^.* at //")
 316         vmdown=$(grep "doctor mark vm.* error at" inspector.log |tail -n 1 |\
 317                  sed -e "s/^.* at //")
 318         hostdown=$(grep "doctor mark host.* down at" inspector.log |\
 319                  sed -e "s/^.* at //")
 320
 321         # TODO(yujunz) check the actual delay to verify time sync status
 322         # expected ~1s delay from $trigger to $linkdown
 323         relative_start=${linkdown}
 324         export DOCTOR_PROFILER_T00=$(python -c \
 325           "print(int(($linkdown-$relative_start)*1000))")
 326         export DOCTOR_PROFILER_T01=$(python -c \
 327           "print(int(($detected-$relative_start)*1000))")
 328         export DOCTOR_PROFILER_T03=$(python -c \
 329           "print(int(($vmdown-$relative_start)*1000))")
 330         export DOCTOR_PROFILER_T04=$(python -c \
 331           "print(int(($hostdown-$relative_start)*1000))")
 332         export DOCTOR_PROFILER_T09=$(python -c \
 333           "print(int(($notified-$relative_start)*1000))")
 334
 335         python profiler-poc.py >doctor_profiler.log 2>&1
 336     fi
 337 }
 338
 339 cleanup() {
 340     set +e
 341     echo "cleanup..."
 342     stop_monitor
 343     stop_inspector
 344     stop_consumer
 345
 346     unset_forced_down_hosts
 347     # TODO: We need to make sure the target compute host is back to IP
 348     #       reachable. wait_ping() will be added by tojuvone .
 349     sleep 110
 350     scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
 351
 352     openstack $as_doctor_user server list | grep -q " $VM_NAME " && openstack $as_doctor_user server delete "$VM_NAME"
 353     sleep 1
 354     alarm_id=$(ceilometer $as_doctor_user alarm-list | grep " $ALARM_NAME " | awk '{print $2}')
 355     sleep 1
 356     [ -n "$alarm_id" ] && ceilometer $as_doctor_user alarm-delete "$alarm_id"
 357     sleep 1
 358
 359     image_id=$(openstack image list | grep " $IMAGE_NAME " | awk '{print $2}')
 360     sleep 1
 361     #if an existing image was used, there's no need to remove it here
 362     if [[ "$use_existing_image" == false ]] ; then
 363         [ -n "$image_id" ] && openstack image delete "$image_id"
 364     fi
 365     openstack role remove "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 366                               --project "$DOCTOR_PROJECT"
 367     openstack project delete "$DOCTOR_PROJECT"
 368     openstack user delete "$DOCTOR_USER"
 369
 370     cleanup_installer
 371     cleanup_inspector
 372 }
 373
 374 # Main process
 375
 376 echo "Note: doctor/tests/run.sh has been executed."
 377 git log --oneline -1 || true   # ignore even you don't have git installed
 378
 379 trap cleanup EXIT
 380
 381 source $TOP_DIR/functions-common
 382 source $TOP_DIR/lib/installer
 383 source $TOP_DIR/lib/inspector
 384
 385 setup_installer
 386
 387 echo "preparing VM image..."
 388 download_image
 389 register_image
 390
 391 echo "creating test user..."
 392 create_test_user
 393
 394 echo "creating VM..."
 395 boot_vm
 396 wait_for_vm_launch
 397
 398 echo "get computer host info..."
 399 get_compute_host_info
 400
 401 echo "creating alarm..."
 402 #TODO: change back to use, network problems depends on infra and installers
 403 #get_consumer_ip
 404 create_alarm
 405
 406 echo "starting doctor sample components..."
 407 start_inspector
 408 start_monitor
 409 start_consumer
 410
 411 sleep 60
 412 echo "injecting host failure..."
 413 inject_failure
 414 sleep 60
 415
 416 check_host_status "(DOWN|UNKNOWN)"
 417 calculate_notification_time
 418 collect_logs
 419 run_profiler
 420
 421 echo "done"