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 DOCTOR_ROLE=_member_
  32 PROFILER_TYPE=${PROFILER_TYPE:-poc}
  33 PYTHON_ENABLE=${PYTHON_ENABLE:-false}
  34
  35 TOP_DIR=$(cd $(dirname "$0") && pwd)
  36
  37 as_doctor_user="--os-username $DOCTOR_USER --os-password $DOCTOR_PW
  38                 --os-project-name $DOCTOR_PROJECT --os-tenant-name $DOCTOR_PROJECT"
  39 # NOTE: ceilometer command still requires '--os-tenant-name'.
  40 #ceilometer="ceilometer ${as_doctor_user/--os-project-name/--os-tenant-name}"
  41 ceilometer="ceilometer $as_doctor_user"
  42 as_admin_user="--os-username admin --os-project-name $DOCTOR_PROJECT
  43                --os-tenant-name $DOCTOR_PROJECT"
  44
  45
  46 # Functions
  47
  48 get_compute_host_info() {
  49     # get computer host info which first VM boot in as admin user
  50     COMPUTE_HOST=$(openstack $as_admin_user server show ${VM_BASENAME}1 |
  51                    grep "OS-EXT-SRV-ATTR:host " | awk '{ print $4 }')
  52     compute_host_in_undercloud=${COMPUTE_HOST%%.*}
  53     die_if_not_set $LINENO COMPUTE_HOST "Failed to get compute hostname"
  54
  55     get_compute_ip_from_hostname $COMPUTE_HOST
  56
  57     echo "COMPUTE_HOST=$COMPUTE_HOST"
  58     echo "COMPUTE_IP=$COMPUTE_IP"
  59
  60     # verify connectivity to target compute host
  61     ping -c 1 "$COMPUTE_IP"
  62     if [[ $? -ne 0 ]] ; then
  63         die $LINENO "Can not ping to computer host"
  64     fi
  65
  66     # verify ssh to target compute host
  67     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'exit'
  68     if [[ $? -ne 0 ]] ; then
  69         die $LINENO "Can not ssh to computer host"
  70     fi
  71 }
  72
  73 # TODO(r-mibu): update this function to support consumer instance
  74 #               and migrate this function into installer lib
  75 get_consumer_ip___to_be_removed() {
  76     local get_consumer_command="ip route get $COMPUTE_IP | awk '/ src /{print \$NF}'"
  77     if is_installer apex; then
  78         CONSUMER_IP=$(sudo ssh $ssh_opts root@$INSTALLER_IP \
  79                       "$get_consumer_command")
  80     elif is_installer fuel; then
  81         CONSUMER_IP=$(sudo sshpass -p r00tme ssh $ssh_opts root@${INSTALLER_IP} \
  82                       "$get_consumer_command")
  83     elif is_installer local; then
  84         CONSUMER_IP=`$get_consumer_command`
  85     fi
  86     echo "CONSUMER_IP=$CONSUMER_IP"
  87
  88     die_if_not_set $LINENO CONSUMER_IP "Could not get CONSUMER_IP."
  89 }
  90
  91 download_image() {
  92     #if a different name was provided for the image in the enviroment there's no need to download the image
  93     use_existing_image=false
  94     openstack image list | grep -q " $IMAGE_NAME " && use_existing_image=true
  95
  96     if [[ "$use_existing_image" == false ]] ; then
  97         [ -e "$IMAGE_FILE" ] && return 0
  98         wget "$IMAGE_URL" -o "$IMAGE_FILE"
  99     fi
 100 }
 101
 102 register_image() {
 103     openstack image list | grep -q " $IMAGE_NAME " && return 0
 104     openstack image create "$IMAGE_NAME" \
 105                            --public \
 106                            --disk-format "$IMAGE_FORMAT" \
 107                            --container-format bare \
 108                            --file "$IMAGE_FILE"
 109 }
 110
 111 create_test_user() {
 112     openstack project list | grep -q " $DOCTOR_PROJECT " || {
 113         openstack project create --description "Doctor Project" \
 114                                  "$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 role show "$DOCTOR_ROLE" | grep -q " $DOCTOR_ROLE " || {
 121         openstack role create "$DOCTOR_ROLE"
 122     }
 123     openstack role assignment list --user "$DOCTOR_USER" \
 124     --project "$DOCTOR_PROJECT" --names | grep -q " $DOCTOR_ROLE " || {
 125         openstack role add "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 126                            --project "$DOCTOR_PROJECT"
 127     }
 128     openstack role assignment list --user admin --project "$DOCTOR_PROJECT" \
 129     --names | grep -q " admin " || {
 130         openstack role add admin --user admin --project "$DOCTOR_PROJECT"
 131     }
 132     # tojuvone: openstack quota show is broken and have to use nova
 133     # https://bugs.launchpad.net/manila/+bug/1652118
 134     # Note! while it is encouraged to use openstack client it has proven
 135     # quite buggy.
 136     # QUOTA=$(openstack quota show $DOCTOR_PROJECT)
 137     DOCTOR_QUOTA=$(nova quota-show --tenant $DOCTOR_PROJECT)
 138     # We make sure that quota allows number of instances and cores
 139     OLD_INSTANCE_QUOTA=$(echo "${DOCTOR_QUOTA}" | grep " instances " | \
 140                          awk '{print $4}')
 141     if [ $OLD_INSTANCE_QUOTA -lt $VM_COUNT ]; then
 142         openstack quota set --instances $VM_COUNT \
 143                   $DOCTOR_USER
 144     fi
 145     OLD_CORES_QUOTA=$(echo "${DOCTOR_QUOTA}" | grep " cores " | \
 146                       awk '{print $4}')
 147     if [ $OLD_CORES_QUOTA -lt $VM_COUNT ]; then
 148         openstack quota set --cores $VM_COUNT \
 149                   $DOCTOR_USER
 150     fi
 151 }
 152
 153 remove_test_user() {
 154     openstack project list | grep -q " $DOCTOR_PROJECT " && {
 155         openstack role assignment list --user admin \
 156         --project "$DOCTOR_PROJECT" --names | grep -q " admin " && {
 157             openstack role remove admin --user admin --project "$DOCTOR_PROJECT"
 158         }
 159         openstack user list | grep -q " $DOCTOR_USER " && {
 160             openstack role assignment list --user "$DOCTOR_USER" \
 161             --project "$DOCTOR_PROJECT" --names | grep -q " $DOCTOR_ROLE " && {
 162                 openstack role remove "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 163                 --project "$DOCTOR_PROJECT"
 164             }
 165             openstack user delete "$DOCTOR_USER"
 166         }
 167         openstack project delete "$DOCTOR_PROJECT"
 168     }
 169 }
 170
 171 boot_vm() {
 172     # test VM done with test user, so can test non-admin
 173
 174     if ! openstack $as_doctor_user network show $NET_NAME; then
 175         openstack $as_doctor_user network create $NET_NAME
 176     fi
 177     if ! openstack $as_doctor_user subnet show $NET_NAME; then
 178         openstack $as_doctor_user subnet create $NET_NAME \
 179             --network $NET_NAME --subnet-range $NET_CIDR --no-dhcp
 180     fi
 181     net_id=$(openstack $as_doctor_user network show $NET_NAME -f value -c id)
 182
 183     servers=$(openstack $as_doctor_user server list)
 184     for i in `seq $VM_COUNT`; do
 185         echo "${servers}" | grep -q " $VM_BASENAME$i " && continue
 186         openstack $as_doctor_user server create --flavor "$VM_FLAVOR" \
 187             --image "$IMAGE_NAME" --nic net-id=$net_id "$VM_BASENAME$i"
 188     done
 189     sleep 1
 190 }
 191
 192 create_alarm() {
 193     # get vm_id as test user
 194     alarm_list=$($ceilometer alarm-list)
 195     vms=$(openstack $as_doctor_user server list)
 196     for i in `seq $VM_COUNT`; do
 197         echo "${alarm_list}" | grep -q " $ALARM_BASENAME$i " || {
 198             vm_id=$(echo "${vms}" | grep " $VM_BASENAME$i " | awk '{print $2}')
 199             # TODO(r-mibu): change notification endpoint from localhost to the
 200             # consumer. IP address (functest container).
 201             $ceilometer alarm-event-create \
 202                        --name "$ALARM_BASENAME$i" \
 203                        --alarm-action "http://localhost:$CONSUMER_PORT/failure" \
 204                        --description "VM failure" \
 205                        --enabled True \
 206                        --repeat-actions False \
 207                        --severity "moderate" \
 208                        --event-type compute.instance.update \
 209                        -q "traits.state=string::error; \
 210                        traits.instance_id=string::$vm_id"
 211             }
 212      done
 213 }
 214
 215 start_consumer() {
 216     pgrep -f "python consumer.py" && return 0
 217     python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 &
 218
 219     # NOTE(r-mibu): create tunnel to the controller nodes, so that we can
 220     # avoid some network problems dpends on infra and installers.
 221     # This tunnel will be terminated by stop_consumer() or after 10 mins passed.
 222     if ! is_installer local; then
 223         for ip in $CONTROLLER_IPS
 224         do
 225             forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
 226             tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
 227             $tunnel_command > "ssh_tunnel.${ip}.log" 2>&1 < /dev/null &
 228         done
 229     fi
 230 }
 231
 232 stop_consumer() {
 233     pgrep -f "python consumer.py" || return 0
 234     kill $(pgrep -f "python consumer.py")
 235
 236     # NOTE(r-mibu): terminate tunnels to the controller nodes
 237     if ! is_installer local; then
 238         for ip in $CONTROLLER_IPS
 239         do
 240             forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
 241             tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
 242             kill $(pgrep -f "$tunnel_command")
 243         done
 244     fi
 245 }
 246
 247 wait_for_vm_launch() {
 248     echo "waiting for vm launch..."
 249
 250     count=0
 251     while [[ ${count} -lt 60 ]]
 252     do
 253         active_count=0
 254         vms=$(openstack $as_doctor_user server list)
 255         for i in `seq $VM_COUNT`; do
 256             state=$(echo "${vms}" | grep " $VM_BASENAME$i " | awk '{print $6}')
 257             if [[ "$state" == "ACTIVE" ]]; then
 258                 active_count=$(($active_count+1))
 259             elif [[ "$state" == "ERROR" ]]; then
 260                 die $LINENO "vm state $VM_BASENAME$i is ERROR"
 261             else
 262                 #This VM not yet active
 263                 count=$(($count+1))
 264                 sleep 5
 265                 continue
 266             fi
 267         done
 268         [[ $active_count -eq $VM_COUNT ]] && {
 269             echo "get computer host info..."
 270             get_compute_host_info
 271             VMS_ON_FAILED_HOST=$(openstack $as_doctor_user server list --host \
 272                          $COMPUTE_HOST | grep " ${VM_BASENAME}" |  wc -l)
 273             return 0
 274         }
 275         #Not all VMs active
 276         count=$(($count+1))
 277         sleep 5
 278     done
 279     die $LINENO "Time out while waiting for VM launch"
 280 }
 281
 282 inject_failure() {
 283     echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..."
 284     cat > disable_network.sh << 'END_TXT'
 285 #!/bin/bash -x
 286 sleep 1
 287 if [ -n "@INTERFACE_NAME@" ]; then
 288     dev=@INTERFACE_NAME@
 289 else
 290     dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $NF}')
 291 fi
 292 sudo ip link set $dev down
 293 echo "doctor set link down at" $(date "+%s.%N")
 294 sleep 180
 295 sudo ip link set $dev up
 296 sleep 1
 297 END_TXT
 298     sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh
 299     sed -i -e "s/@INTERFACE_NAME@/$INTERFACE_NAME/" disable_network.sh
 300     chmod +x disable_network.sh
 301     scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:"
 302     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &'
 303     # use host time to get rid of potential time sync deviation between nodes
 304     triggered=$(date "+%s.%N")
 305 }
 306
 307 wait_consumer() {
 308     local interval=1
 309     local rounds=$(($1 / $interval))
 310     for i in `seq $rounds`; do
 311         notified_count=$(grep "doctor consumer notified at" consumer.log | wc -l)
 312         if [[ $notified_count -eq  $VMS_ON_FAILED_HOST ]]; then
 313             return 0
 314         fi
 315         sleep $interval
 316     done
 317     die $LINENO "Consumer hasn't received fault notification."
 318 }
 319
 320 calculate_notification_time() {
 321     wait_consumer 60
 322     #keep 'at' as the last keyword just before the value, and
 323     #use regex to get value instead of the fixed column
 324     if [ ! -f monitor.log ]; then
 325         scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:monitor.log" .
 326     fi
 327     detected=$(grep "doctor monitor detected at" monitor.log |\
 328                sed -e "s/^.* at //" | tail -1)
 329     notified=$(grep "doctor consumer notified at" consumer.log |\
 330                sed -e "s/^.* at //" | tail -1)
 331
 332     echo "$notified $detected" | \
 333         awk '{
 334             d = $1 - $2;
 335             if (d < 1 && d > 0) { print d " OK"; exit 0 }
 336             else { print d " NG"; exit 1 }
 337         }'
 338 }
 339
 340 check_host_status() {
 341     # Check host related to first Doctor VM is in wanted state
 342     # $1    Expected state
 343     # $2    Seconds to wait to have wanted state
 344     expected_state=$1
 345     local interval=5
 346     local rounds=$(($2 / $interval))
 347     for i in `seq $rounds`; do
 348         host_status_line=$(openstack $as_doctor_user --os-compute-api-version \
 349                            2.16 server show ${VM_BASENAME}1 | grep "host_status")
 350         host_status=$(echo $host_status_line | awk '{print $4}')
 351         die_if_not_set $LINENO host_status "host_status not reported by: nova show ${VM_BASENAME}1"
 352         if [[ "$expected_state" =~ "$host_status" ]] ; then
 353             echo "${VM_BASENAME}1 showing host_status: $host_status"
 354             return 0
 355         else
 356             sleep $interval
 357         fi
 358     done
 359     if [[ "$expected_state" =~ "$host_status" ]] ; then
 360         echo "${VM_BASENAME}1 showing host_status: $host_status"
 361     else
 362         die $LINENO  "host_status:$host_status not equal to expected_state: $expected_state"
 363     fi
 364 }
 365
 366 unset_forced_down_hosts() {
 367     # for debug
 368     openstack compute service list --service nova-compute
 369
 370     downed_computes=$(openstack compute service list --service nova-compute \
 371                       -f value -c Host -c State | grep ' down$' \
 372                       | sed -e 's/ *down$//')
 373     echo "downed_computes: $downed_computes"
 374     for host in $downed_computes
 375     do
 376         # TODO(r-mibu): use openstack client
 377         #openstack compute service set --up $host nova-compute
 378         nova service-force-down --unset $host nova-compute
 379     done
 380
 381     echo "waiting disabled compute host back to be enabled..."
 382     wait_until 'openstack compute service list --service nova-compute
 383                 -f value -c State | grep -q down' 240 5
 384
 385     for host in $downed_computes
 386     do
 387         # TODO(r-mibu): improve 'get_compute_ip_from_hostname'
 388         get_compute_ip_from_hostname $host
 389         wait_until "! ping -c 1 $COMPUTE_IP" 120 5
 390     done
 391 }
 392
 393 collect_logs() {
 394     if [[ -n "$COMPUTE_IP" ]];then
 395         scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
 396     fi
 397
 398     # TODO(yujunz) collect other logs, e.g. nova, aodh
 399 }
 400
 401 run_profiler() {
 402     if [[ "$PROFILER_TYPE" == "poc" ]]; then
 403         linkdown=$(grep "doctor set link down at " disable_network.log |\
 404                   sed -e "s/^.* at //")
 405         vmdown=$(grep "doctor mark vm.* error at" inspector.log |tail -n 1 |\
 406                  sed -e "s/^.* at //")
 407         hostdown=$(grep "doctor mark host.* down at" inspector.log |\
 408                  sed -e "s/^.* at //")
 409
 410         # TODO(yujunz) check the actual delay to verify time sync status
 411         # expected ~1s delay from $trigger to $linkdown
 412         relative_start=${linkdown}
 413         export DOCTOR_PROFILER_T00=$(python -c \
 414           "print(int(($linkdown-$relative_start)*1000))")
 415         export DOCTOR_PROFILER_T01=$(python -c \
 416           "print(int(($detected-$relative_start)*1000))")
 417         export DOCTOR_PROFILER_T03=$(python -c \
 418           "print(int(($vmdown-$relative_start)*1000))")
 419         export DOCTOR_PROFILER_T04=$(python -c \
 420           "print(int(($hostdown-$relative_start)*1000))")
 421         export DOCTOR_PROFILER_T09=$(python -c \
 422           "print(int(($notified-$relative_start)*1000))")
 423
 424         python profiler_poc.py > doctor_profiler.log 2>&1
 425     fi
 426 }
 427
 428 cleanup() {
 429     set +e
 430     echo "cleanup..."
 431     stop_inspector
 432     stop_consumer
 433
 434     unset_forced_down_hosts
 435     stop_monitor
 436     collect_logs
 437
 438     vms=$(openstack $as_doctor_user server list)
 439     vmstodel=""
 440     for i in `seq $VM_COUNT`; do
 441         $(echo "${vms}" | grep -q " $VM_BASENAME$i ") &&
 442         vmstodel+=" $VM_BASENAME$i"
 443     done
 444     [[ $vmstodel ]] && openstack $as_doctor_user server delete $vmstodel
 445     alarm_list=$($ceilometer alarm-list)
 446     for i in `seq $VM_COUNT`; do
 447         alarm_id=$(echo "${alarm_list}" | grep " $ALARM_BASENAME$i " |
 448                    awk '{print $2}')
 449         [ -n "$alarm_id" ] && $ceilometer alarm-delete "$alarm_id"
 450     done
 451     openstack $as_doctor_user subnet delete $NET_NAME
 452     sleep 1
 453     openstack $as_doctor_user network delete $NET_NAME
 454     sleep 1
 455
 456     image_id=$(openstack image list | grep " $IMAGE_NAME " | awk '{print $2}')
 457     sleep 1
 458     #if an existing image was used, there's no need to remove it here
 459     if [[ "$use_existing_image" == false ]] ; then
 460         [ -n "$image_id" ] && openstack image delete "$image_id"
 461     fi
 462
 463     remove_test_user
 464
 465     cleanup_installer
 466     cleanup_inspector
 467     cleanup_monitor
 468
 469     # NOTE: Temporal log printer.
 470     for f in $(find . -name '*.log')
 471     do
 472         echo
 473         echo "[$f]"
 474         sed -e 's/^/ | /' $f
 475         echo
 476     done
 477 }
 478
 479 setup_python_packages() {
 480     sudo pip install flask==0.10.1
 481     command -v openstack || sudo pip install python-openstackclient==2.3.0
 482     command -v ceilometer || sudo pip install python-ceilometerclient==2.6.2
 483     command -v congress || sudo pip install python-congressclient==1.5.0
 484 }
 485
 486 # Main process
 487
 488 if [[ $PYTHON_ENABLE == [Tt]rue ]]; then
 489     which tox || sudo pip install tox
 490     if [ -f /usr/bin/apt-get ]; then
 491         sudo apt-get install -y python3-dev
 492     elif [ -f /usr/bin/yum ] ; then
 493         sudo yum install -y python3-devel
 494     fi
 495
 496     cd $TOP_DIR
 497     echo "executing tox..."
 498     tox
 499     exit $?
 500 fi
 501
 502 echo "Note: doctor/tests/run.sh has been executed."
 503 git log --oneline -1 || true   # ignore even you don't have git installed
 504
 505 trap cleanup EXIT
 506
 507 setup_python_packages
 508
 509 source $TOP_DIR/functions-common
 510 source $TOP_DIR/lib/installer
 511 source $TOP_DIR/lib/inspector
 512 source $TOP_DIR/lib/monitor
 513
 514 rm -f *.log
 515
 516 setup_installer
 517
 518 echo "preparing VM image..."
 519 download_image
 520 register_image
 521
 522 echo "creating test user..."
 523 create_test_user
 524
 525 echo "creating VM..."
 526 boot_vm
 527 wait_for_vm_launch
 528
 529 echo "creating alarm..."
 530 #TODO: change back to use, network problems depends on infra and installers
 531 #get_consumer_ip
 532 create_alarm
 533
 534 echo "starting doctor sample components..."
 535 start_inspector
 536 start_monitor
 537 start_consumer
 538
 539 sleep 60
 540 echo "injecting host failure..."
 541 inject_failure
 542
 543 check_host_status "(DOWN|UNKNOWN)" 60
 544 unset_forced_down_hosts
 545 calculate_notification_time
 546 collect_logs
 547 run_profiler
 548
 549 echo "done"