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="${IMAGE_NAME}.img"
  19 IMAGE_FORMAT=qcow2
  20 VM_BASENAME=doctor_vm
  21 VM_FLAVOR=m1.tiny
  22 #if VM_COUNT set, use that instead
  23 VM_COUNT=${VM_COUNT:-1}
  24 NET_NAME=doctor_net
  25 NET_CIDR=192.168.168.0/24
  26 ALARM_BASENAME=doctor_alarm
  27 CONSUMER_PORT=12346
  28 DOCTOR_USER=doctor
  29 DOCTOR_PW=doctor
  30 DOCTOR_PROJECT=doctor
  31 #TODO: change back to `_member_` when JIRA DOCTOR-55 is done
  32 DOCTOR_ROLE=admin
  33 PROFILER_TYPE=${PROFILER_TYPE:-none}
  34 PYTHON_ENABLE=${PYTHON_ENABLE:-false}
  35
  36 TOP_DIR=$(cd $(dirname "$0") && pwd)
  37
  38 as_doctor_user="--os-username $DOCTOR_USER --os-password $DOCTOR_PW
  39                 --os-project-name $DOCTOR_PROJECT --os-tenant-name $DOCTOR_PROJECT"
  40 # NOTE: ceilometer command still requires '--os-tenant-name'.
  41 #ceilometer="ceilometer ${as_doctor_user/--os-project-name/--os-tenant-name}"
  42 ceilometer="ceilometer $as_doctor_user"
  43
  44
  45 # Functions
  46
  47 get_compute_host_info() {
  48     # get computer host info which first VM boot in
  49     COMPUTE_HOST=$(openstack $as_doctor_user server show ${VM_BASENAME}1 |
  50                    grep "OS-EXT-SRV-ATTR:host" | awk '{ print $4 }')
  51     compute_host_in_undercloud=${COMPUTE_HOST%%.*}
  52     die_if_not_set $LINENO COMPUTE_HOST "Failed to get compute hostname"
  53
  54     get_compute_ip_from_hostname $COMPUTE_HOST
  55
  56     echo "COMPUTE_HOST=$COMPUTE_HOST"
  57     echo "COMPUTE_IP=$COMPUTE_IP"
  58
  59     # verify connectivity to target compute host
  60     ping -c 1 "$COMPUTE_IP"
  61     if [[ $? -ne 0 ]] ; then
  62         die $LINENO "Can not ping to computer host"
  63     fi
  64
  65     # verify ssh to target compute host
  66     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'exit'
  67     if [[ $? -ne 0 ]] ; then
  68         die $LINENO "Can not ssh to computer host"
  69     fi
  70 }
  71
  72 # TODO(r-mibu): update this function to support consumer instance
  73 #               and migrate this function into installer lib
  74 get_consumer_ip___to_be_removed() {
  75     local get_consumer_command="ip route get $COMPUTE_IP | awk '/ src /{print \$NF}'"
  76     if is_installer apex; then
  77         CONSUMER_IP=$(sudo ssh $ssh_opts root@$INSTALLER_IP \
  78                       "$get_consumer_command")
  79     elif is_installer fuel; then
  80         CONSUMER_IP=$(sudo sshpass -p r00tme ssh $ssh_opts root@${INSTALLER_IP} \
  81                       "$get_consumer_command")
  82     elif is_installer local; then
  83         CONSUMER_IP=`$get_consumer_command`
  84     fi
  85     echo "CONSUMER_IP=$CONSUMER_IP"
  86
  87     die_if_not_set $LINENO CONSUMER_IP "Could not get CONSUMER_IP."
  88 }
  89
  90 download_image() {
  91     #if a different name was provided for the image in the enviroment there's no need to download the image
  92     use_existing_image=false
  93     openstack image list | grep -q " $IMAGE_NAME " && use_existing_image=true
  94
  95     if [[ "$use_existing_image" == false ]] ; then
  96         [ -e "$IMAGE_FILE" ] && return 0
  97         wget "$IMAGE_URL" -o "$IMAGE_FILE"
  98     fi
  99 }
 100
 101 register_image() {
 102     openstack image list | grep -q " $IMAGE_NAME " && return 0
 103     openstack image create "$IMAGE_NAME" \
 104                            --public \
 105                            --disk-format "$IMAGE_FORMAT" \
 106                            --container-format bare \
 107                            --file "$IMAGE_FILE"
 108 }
 109
 110 create_test_user() {
 111     openstack project list | grep -q " $DOCTOR_PROJECT " || {
 112         openstack project create "$DOCTOR_PROJECT"
 113     }
 114     openstack user list | grep -q " $DOCTOR_USER " || {
 115         openstack user create "$DOCTOR_USER" --password "$DOCTOR_PW" \
 116                               --project "$DOCTOR_PROJECT"
 117     }
 118     openstack role show "$DOCTOR_ROLE" || {
 119         openstack role create "$DOCTOR_ROLE"
 120     }
 121     openstack role add "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 122                        --project "$DOCTOR_PROJECT"
 123     # tojuvone: openstack quota show is broken and have to use nova
 124     # https://bugs.launchpad.net/manila/+bug/1652118
 125     # Note! while it is encouraged to use openstack client it has proven
 126     # quite buggy.
 127     # QUOTA=$(openstack quota show $DOCTOR_PROJECT)
 128     DOCTOR_QUOTA=$(nova quota-show --tenant $DOCTOR_PROJECT)
 129     # We make sure that quota allows number of instances and cores
 130     OLD_INSTANCE_QUOTA=$(echo "${DOCTOR_QUOTA}" | grep " instances " | \
 131                          awk '{print $4}')
 132     if [ $OLD_INSTANCE_QUOTA -lt $VM_COUNT ]; then
 133         openstack quota set --instances $VM_COUNT \
 134                   $DOCTOR_USER
 135     fi
 136     OLD_CORES_QUOTA=$(echo "${DOCTOR_QUOTA}" | grep " cores " | \
 137                       awk '{print $4}')
 138     if [ $OLD_CORES_QUOTA -lt $VM_COUNT ]; then
 139         openstack quota set --cores $VM_COUNT \
 140                   $DOCTOR_USER
 141     fi
 142 }
 143
 144 boot_vm() {
 145     # test VM done with test user, so can test non-admin
 146
 147     if ! openstack $as_doctor_user network show $NET_NAME; then
 148         openstack $as_doctor_user network create $NET_NAME
 149     fi
 150     if ! openstack $as_doctor_user subnet show $NET_NAME; then
 151         openstack $as_doctor_user subnet create $NET_NAME \
 152             --network $NET_NAME --subnet-range $NET_CIDR --no-dhcp
 153     fi
 154     net_id=$(openstack $as_doctor_user network show $NET_NAME -f value -c id)
 155
 156     servers=$(openstack $as_doctor_user server list)
 157     for i in `seq $VM_COUNT`; do
 158         echo "${servers}" | grep -q " $VM_BASENAME$i " && continue
 159         openstack $as_doctor_user server create --flavor "$VM_FLAVOR" \
 160             --image "$IMAGE_NAME" --nic net-id=$net_id "$VM_BASENAME$i"
 161     done
 162     sleep 1
 163 }
 164
 165 create_alarm() {
 166     # get vm_id as test user
 167     alarm_list=$($ceilometer alarm-list)
 168     vms=$(openstack $as_doctor_user server list)
 169     for i in `seq $VM_COUNT`; do
 170         echo "${alarm_list}" | grep -q " $ALARM_BASENAME$i " || {
 171             vm_id=$(echo "${vms}" | grep " $VM_BASENAME$i " | awk '{print $2}')
 172             # TODO(r-mibu): change notification endpoint from localhost to the
 173             # consumer. IP address (functest container).
 174             $ceilometer alarm-event-create \
 175                        --name "$ALARM_BASENAME$i" \
 176                        --alarm-action "http://localhost:$CONSUMER_PORT/failure" \
 177                        --description "VM failure" \
 178                        --enabled True \
 179                        --repeat-actions False \
 180                        --severity "moderate" \
 181                        --event-type compute.instance.update \
 182                        -q "traits.state=string::error; \
 183                        traits.instance_id=string::$vm_id"
 184             }
 185      done
 186 }
 187
 188 start_monitor() {
 189     pgrep -f "python monitor.py" && return 0
 190     sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
 191         > monitor.log 2>&1 &
 192 }
 193
 194 stop_monitor() {
 195     pgrep -f "python monitor.py" || return 0
 196     sudo kill $(pgrep -f "python monitor.py")
 197 }
 198
 199 start_consumer() {
 200     pgrep -f "python consumer.py" && return 0
 201     python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 &
 202
 203     # NOTE(r-mibu): create tunnel to the controller nodes, so that we can
 204     # avoid some network problems dpends on infra and installers.
 205     # This tunnel will be terminated by stop_consumer() or after 10 mins passed.
 206     if ! is_installer local; then
 207         for ip in $CONTROLLER_IPS
 208         do
 209             forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
 210             tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
 211             $tunnel_command > "ssh_tunnel.${ip}.log" 2>&1 < /dev/null &
 212         done
 213     fi
 214 }
 215
 216 stop_consumer() {
 217     pgrep -f "python consumer.py" || return 0
 218     kill $(pgrep -f "python consumer.py")
 219
 220     # NOTE(r-mibu): terminate tunnels to the controller nodes
 221     if ! is_installer local; then
 222         for ip in $CONTROLLER_IPS
 223         do
 224             forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
 225             tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
 226             kill $(pgrep -f "$tunnel_command")
 227         done
 228     fi
 229 }
 230
 231 wait_for_vm_launch() {
 232     echo "waiting for vm launch..."
 233
 234     count=0
 235     while [[ ${count} -lt 60 ]]
 236     do
 237         active_count=0
 238         vms=$(openstack $as_doctor_user server list)
 239         for i in `seq $VM_COUNT`; do
 240             state=$(echo "${vms}" | grep " $VM_BASENAME$i " | awk '{print $6}')
 241             if [[ "$state" == "ACTIVE" ]]; then
 242                 active_count=$(($active_count+1))
 243             elif [[ "$state" == "ERROR" ]]; then
 244                 die $LINENO "vm state $VM_BASENAME$i is ERROR"
 245             else
 246                 #This VM not yet active
 247                 count=$(($count+1))
 248                 sleep 5
 249                 continue
 250             fi
 251         done
 252         [[ $active_count -eq $VM_COUNT ]] && {
 253             echo "get computer host info..."
 254             get_compute_host_info
 255             VMS_ON_FAILED_HOST=$(openstack $as_doctor_user server list --host \
 256                          $COMPUTE_HOST | grep " ${VM_BASENAME}" |  wc -l)
 257             return 0
 258         }
 259         #Not all VMs active
 260         count=$(($count+1))
 261         sleep 5
 262     done
 263     die $LINENO "Time out while waiting for VM launch"
 264 }
 265
 266 inject_failure() {
 267     echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..."
 268     cat > disable_network.sh << 'END_TXT'
 269 #!/bin/bash -x
 270 dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $7}')
 271 [[ -n "$dev" ]] || dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $5}')
 272 sleep 1
 273 sudo ip link set $dev down
 274 echo "doctor set link down at" $(date "+%s.%N")
 275 sleep 180
 276 sudo ip link set $dev up
 277 sleep 1
 278 END_TXT
 279     sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh
 280     chmod +x disable_network.sh
 281     scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:"
 282     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &'
 283     # use host time to get rid of potential time sync deviation between nodes
 284     triggered=$(date "+%s.%N")
 285 }
 286
 287 wait_consumer() {
 288     local interval=1
 289     local rounds=$(($1 / $interval))
 290     for i in `seq $rounds`; do
 291         notified_count=$(grep "doctor consumer notified at" consumer.log | wc -l)
 292         if [[ $notified_count -eq  $VMS_ON_FAILED_HOST ]]; then
 293             return 0
 294         fi
 295         sleep $interval
 296     done
 297     die $LINENO "Consumer hasn't received fault notification."
 298 }
 299
 300 calculate_notification_time() {
 301     wait_consumer 60
 302     #keep 'at' as the last keyword just before the value, and
 303     #use regex to get value instead of the fixed column
 304     detected=$(grep "doctor monitor detected at" monitor.log |\
 305                sed -e "s/^.* at //")
 306     notified=$(grep "doctor consumer notified at" consumer.log |\
 307                sed -e "s/^.* at //" | tail -1)
 308
 309     echo "$notified $detected" | \
 310         awk '{
 311             d = $1 - $2;
 312             if (d < 1 && d > 0) { print d " OK"; exit 0 }
 313             else { print d " NG"; exit 1 }
 314         }'
 315 }
 316
 317 check_host_status() {
 318     # Check host related to first Doctor VM is in wanted state
 319     # $1    Expected state
 320     # $2    Seconds to wait to have wanted state
 321     expected_state=$1
 322     local interval=5
 323     local rounds=$(($2 / $interval))
 324     for i in `seq $rounds`; do
 325         host_status_line=$(openstack $as_doctor_user --os-compute-api-version \
 326                            2.16 server show ${VM_BASENAME}1 | grep "host_status")
 327         host_status=$(echo $host_status_line | awk '{print $4}')
 328         die_if_not_set $LINENO host_status "host_status not reported by: nova show ${VM_BASENAME}1"
 329         if [[ "$expected_state" =~ "$host_status" ]] ; then
 330             echo "${VM_BASENAME}1 showing host_status: $host_status"
 331             return 0
 332         else
 333             sleep $interval
 334         fi
 335     done
 336     if [[ "$expected_state" =~ "$host_status" ]] ; then
 337         echo "${VM_BASENAME}1 showing host_status: $host_status"
 338     else
 339         die $LINENO  "host_status:$host_status not equal to expected_state: $expected_state"
 340     fi
 341 }
 342
 343 unset_forced_down_hosts() {
 344     # for debug
 345     openstack compute service list --service nova-compute
 346
 347     downed_computes=$(openstack compute service list --service nova-compute \
 348                       -f value -c Host -c State | grep ' down$' \
 349                       | sed -e 's/ *down$//')
 350     echo "downed_computes: $downed_computes"
 351     for host in $downed_computes
 352     do
 353         # TODO(r-mibu): use openstack client
 354         #openstack compute service set --up $host nova-compute
 355         nova service-force-down --unset $host nova-compute
 356     done
 357
 358     echo "waiting disabled compute host back to be enabled..."
 359     wait_until 'openstack compute service list --service nova-compute
 360                 -f value -c State | grep -q down' 240 5
 361
 362     for host in $downed_computes
 363     do
 364         # TODO(r-mibu): improve 'get_compute_ip_from_hostname'
 365         get_compute_ip_from_hostname $host
 366         wait_until "! ping -c 1 $COMPUTE_IP" 120 5
 367     done
 368 }
 369
 370 collect_logs() {
 371     if [[ -n "$COMPUTE_IP" ]];then
 372         scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
 373     fi
 374
 375     # TODO(yujunz) collect other logs, e.g. nova, aodh
 376 }
 377
 378 run_profiler() {
 379     if [[ "$PROFILER_TYPE" == "poc" ]]; then
 380         linkdown=$(grep "doctor set link down at " disable_network.log |\
 381                   sed -e "s/^.* at //")
 382         vmdown=$(grep "doctor mark vm.* error at" inspector.log |tail -n 1 |\
 383                  sed -e "s/^.* at //")
 384         hostdown=$(grep "doctor mark host.* down at" inspector.log |\
 385                  sed -e "s/^.* at //")
 386
 387         # TODO(yujunz) check the actual delay to verify time sync status
 388         # expected ~1s delay from $trigger to $linkdown
 389         relative_start=${linkdown}
 390         export DOCTOR_PROFILER_T00=$(python -c \
 391           "print(int(($linkdown-$relative_start)*1000))")
 392         export DOCTOR_PROFILER_T01=$(python -c \
 393           "print(int(($detected-$relative_start)*1000))")
 394         export DOCTOR_PROFILER_T03=$(python -c \
 395           "print(int(($vmdown-$relative_start)*1000))")
 396         export DOCTOR_PROFILER_T04=$(python -c \
 397           "print(int(($hostdown-$relative_start)*1000))")
 398         export DOCTOR_PROFILER_T09=$(python -c \
 399           "print(int(($notified-$relative_start)*1000))")
 400
 401         python profiler-poc.py >doctor_profiler.log 2>&1
 402     fi
 403 }
 404
 405 cleanup() {
 406     set +e
 407     echo "cleanup..."
 408     stop_monitor
 409     stop_inspector
 410     stop_consumer
 411
 412     unset_forced_down_hosts
 413     collect_logs
 414
 415     vms=$(openstack $as_doctor_user server list)
 416     vmstodel=""
 417     for i in `seq $VM_COUNT`; do
 418         $(echo "${vms}" | grep -q " $VM_BASENAME$i ") &&
 419         vmstodel+=" $VM_BASENAME$i"
 420     done
 421     [[ $vmstodel ]] && openstack $as_doctor_user server delete $vmstodel
 422     alarm_list=$($ceilometer alarm-list)
 423     for i in `seq $VM_COUNT`; do
 424         alarm_id=$(echo "${alarm_list}" | grep " $ALARM_BASENAME$i " |
 425                    awk '{print $2}')
 426         [ -n "$alarm_id" ] && $ceilometer alarm-delete "$alarm_id"
 427     done
 428     openstack $as_doctor_user subnet delete $NET_NAME
 429     sleep 1
 430     openstack $as_doctor_user network delete $NET_NAME
 431     sleep 1
 432
 433     image_id=$(openstack image list | grep " $IMAGE_NAME " | awk '{print $2}')
 434     sleep 1
 435     #if an existing image was used, there's no need to remove it here
 436     if [[ "$use_existing_image" == false ]] ; then
 437         [ -n "$image_id" ] && openstack image delete "$image_id"
 438     fi
 439     openstack role remove "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 440                               --project "$DOCTOR_PROJECT"
 441     openstack project delete "$DOCTOR_PROJECT"
 442     openstack user delete "$DOCTOR_USER"
 443     # NOTE: remove role only for doctor test.
 444     #openstack role delete "$DOCTOR_ROLE"
 445
 446     cleanup_installer
 447     cleanup_inspector
 448
 449     # NOTE: Temporal log printer.
 450     for f in $(find . -name '*.log')
 451     do
 452         echo
 453         echo "[$f]"
 454         sed -e 's/^/ | /' $f
 455         echo
 456     done
 457 }
 458
 459 # Main process
 460
 461 if $PYTHON_ENABLE; then
 462     cd $TOP_DIR
 463     echo "executing tox..."
 464     tox
 465     exit $?
 466 fi
 467
 468 echo "Note: doctor/tests/run.sh has been executed."
 469 git log --oneline -1 || true   # ignore even you don't have git installed
 470
 471 trap cleanup EXIT
 472
 473 source $TOP_DIR/functions-common
 474 source $TOP_DIR/lib/installer
 475 source $TOP_DIR/lib/inspector
 476
 477 setup_installer
 478
 479 echo "preparing VM image..."
 480 download_image
 481 register_image
 482
 483 echo "creating test user..."
 484 create_test_user
 485
 486 echo "creating VM..."
 487 boot_vm
 488 wait_for_vm_launch
 489
 490 echo "creating alarm..."
 491 #TODO: change back to use, network problems depends on infra and installers
 492 #get_consumer_ip
 493 create_alarm
 494
 495 echo "starting doctor sample components..."
 496 start_inspector
 497 start_monitor
 498 start_consumer
 499
 500 sleep 60
 501 echo "injecting host failure..."
 502 inject_failure
 503
 504 check_host_status "(DOWN|UNKNOWN)" 60
 505 calculate_notification_time
 506 unset_forced_down_hosts
 507 collect_logs
 508 run_profiler
 509
 510 echo "done"