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=${PROFILER_TYPE:-none}
  33
  34 TOP_DIR=$(cd $(dirname "$0") && pwd)
  35
  36 as_doctor_user="--os-username $DOCTOR_USER --os-password $DOCTOR_PW
  37                 --os-project-name $DOCTOR_PROJECT --os-tenant-name $DOCTOR_PROJECT"
  38 # NOTE: ceilometer command still requires '--os-tenant-name'.
  39 #ceilometer="ceilometer ${as_doctor_user/--os-project-name/--os-tenant-name}"
  40 ceilometer="ceilometer $as_doctor_user"
  41 as_admin_user="--os-username admin --os-project-name $DOCTOR_PROJECT
  42                --os-tenant-name $DOCTOR_PROJECT"
  43
  44
  45 # Functions
  46
  47 get_compute_host_info() {
  48     # get computer host info which first VM boot in as admin user
  49     COMPUTE_HOST=$(openstack $as_admin_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 " $DOCTOR_PROJECT " || {
 112         openstack project create --description "Doctor Project" \
 113                                  "$DOCTOR_PROJECT"
 114     }
 115     openstack user list | grep -q " $DOCTOR_USER " || {
 116         openstack user create "$DOCTOR_USER" --password "$DOCTOR_PW" \
 117                               --project "$DOCTOR_PROJECT"
 118     }
 119     openstack role show "$DOCTOR_ROLE" | grep -q " $DOCTOR_ROLE " || {
 120         openstack role create "$DOCTOR_ROLE"
 121     }
 122     openstack role assignment list --user "$DOCTOR_USER" \
 123     --project "$DOCTOR_PROJECT" --names | grep -q " $DOCTOR_ROLE " || {
 124         openstack role add "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 125                            --project "$DOCTOR_PROJECT"
 126     }
 127     openstack role assignment list --user admin --project "$DOCTOR_PROJECT" \
 128     --names | grep -q " admin " || {
 129         openstack role add admin --user admin --project "$DOCTOR_PROJECT"
 130     }
 131     # tojuvone: openstack quota show is broken and have to use nova
 132     # https://bugs.launchpad.net/manila/+bug/1652118
 133     # Note! while it is encouraged to use openstack client it has proven
 134     # quite buggy.
 135     # QUOTA=$(openstack quota show $DOCTOR_PROJECT)
 136     DOCTOR_QUOTA=$(nova quota-show --tenant $DOCTOR_PROJECT)
 137     # We make sure that quota allows number of instances and cores
 138     OLD_INSTANCE_QUOTA=$(echo "${DOCTOR_QUOTA}" | grep " instances " | \
 139                          awk '{print $4}')
 140     if [ $OLD_INSTANCE_QUOTA -lt $VM_COUNT ]; then
 141         openstack quota set --instances $VM_COUNT \
 142                   $DOCTOR_USER
 143     fi
 144     OLD_CORES_QUOTA=$(echo "${DOCTOR_QUOTA}" | grep " cores " | \
 145                       awk '{print $4}')
 146     if [ $OLD_CORES_QUOTA -lt $VM_COUNT ]; then
 147         openstack quota set --cores $VM_COUNT \
 148                   $DOCTOR_USER
 149     fi
 150 }
 151
 152 remove_test_user() {
 153     openstack project list | grep -q " $DOCTOR_PROJECT " && {
 154         openstack role assignment list --user admin \
 155         --project "$DOCTOR_PROJECT" --names | grep -q " admin " && {
 156             openstack role remove admin --user admin --project "$DOCTOR_PROJECT"
 157         }
 158         openstack user list | grep -q " $DOCTOR_USER " && {
 159             openstack role assignment list --user "$DOCTOR_USER" \
 160             --project "$DOCTOR_PROJECT" --names | grep -q " $DOCTOR_ROLE " && {
 161                 openstack role remove "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 162                 --project "$DOCTOR_PROJECT"
 163             }
 164             openstack user delete "$DOCTOR_USER"
 165         }
 166         openstack project delete "$DOCTOR_PROJECT"
 167     }
 168 }
 169
 170 boot_vm() {
 171     # test VM done with test user, so can test non-admin
 172
 173     if ! openstack $as_doctor_user network show $NET_NAME; then
 174         openstack $as_doctor_user network create $NET_NAME
 175     fi
 176     if ! openstack $as_doctor_user subnet show $NET_NAME; then
 177         openstack $as_doctor_user subnet create $NET_NAME \
 178             --network $NET_NAME --subnet-range $NET_CIDR --no-dhcp
 179     fi
 180     net_id=$(openstack $as_doctor_user network show $NET_NAME -f value -c id)
 181
 182     servers=$(openstack $as_doctor_user server list)
 183     for i in `seq $VM_COUNT`; do
 184         echo "${servers}" | grep -q " $VM_BASENAME$i " && continue
 185         openstack $as_doctor_user server create --flavor "$VM_FLAVOR" \
 186             --image "$IMAGE_NAME" --nic net-id=$net_id "$VM_BASENAME$i"
 187     done
 188     sleep 1
 189 }
 190
 191 create_alarm() {
 192     # get vm_id as test user
 193     alarm_list=$($ceilometer alarm-list)
 194     vms=$(openstack $as_doctor_user server list)
 195     for i in `seq $VM_COUNT`; do
 196         echo "${alarm_list}" | grep -q " $ALARM_BASENAME$i " || {
 197             vm_id=$(echo "${vms}" | grep " $VM_BASENAME$i " | awk '{print $2}')
 198             # TODO(r-mibu): change notification endpoint from localhost to the
 199             # consumer. IP address (functest container).
 200             $ceilometer alarm-event-create \
 201                        --name "$ALARM_BASENAME$i" \
 202                        --alarm-action "http://localhost:$CONSUMER_PORT/failure" \
 203                        --description "VM failure" \
 204                        --enabled True \
 205                        --repeat-actions False \
 206                        --severity "moderate" \
 207                        --event-type compute.instance.update \
 208                        -q "traits.state=string::error; \
 209                        traits.instance_id=string::$vm_id"
 210             }
 211      done
 212 }
 213
 214 start_monitor() {
 215     pgrep -f "python monitor.py" && return 0
 216     sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
 217         > monitor.log 2>&1 &
 218 }
 219
 220 stop_monitor() {
 221     pgrep -f "python monitor.py" || return 0
 222     sudo kill $(pgrep -f "python monitor.py")
 223 }
 224
 225 start_consumer() {
 226     pgrep -f "python consumer.py" && return 0
 227     python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 &
 228
 229     # NOTE(r-mibu): create tunnel to the controller nodes, so that we can
 230     # avoid some network problems dpends on infra and installers.
 231     # This tunnel will be terminated by stop_consumer() or after 10 mins passed.
 232     if ! is_installer local; then
 233         for ip in $CONTROLLER_IPS
 234         do
 235             forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
 236             tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
 237             $tunnel_command > "ssh_tunnel.${ip}.log" 2>&1 < /dev/null &
 238         done
 239     fi
 240 }
 241
 242 stop_consumer() {
 243     pgrep -f "python consumer.py" || return 0
 244     kill $(pgrep -f "python consumer.py")
 245
 246     # NOTE(r-mibu): terminate tunnels to the controller nodes
 247     if ! is_installer local; then
 248         for ip in $CONTROLLER_IPS
 249         do
 250             forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
 251             tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
 252             kill $(pgrep -f "$tunnel_command")
 253         done
 254     fi
 255 }
 256
 257 wait_for_vm_launch() {
 258     echo "waiting for vm launch..."
 259
 260     count=0
 261     while [[ ${count} -lt 60 ]]
 262     do
 263         active_count=0
 264         vms=$(openstack $as_doctor_user server list)
 265         for i in `seq $VM_COUNT`; do
 266             state=$(echo "${vms}" | grep " $VM_BASENAME$i " | awk '{print $6}')
 267             if [[ "$state" == "ACTIVE" ]]; then
 268                 active_count=$(($active_count+1))
 269             elif [[ "$state" == "ERROR" ]]; then
 270                 die $LINENO "vm state $VM_BASENAME$i is ERROR"
 271             else
 272                 #This VM not yet active
 273                 count=$(($count+1))
 274                 sleep 5
 275                 continue
 276             fi
 277         done
 278         [[ $active_count -eq $VM_COUNT ]] && {
 279             echo "get computer host info..."
 280             get_compute_host_info
 281             VMS_ON_FAILED_HOST=$(openstack $as_doctor_user server list --host \
 282                          $COMPUTE_HOST | grep " ${VM_BASENAME}" |  wc -l)
 283             return 0
 284         }
 285         #Not all VMs active
 286         count=$(($count+1))
 287         sleep 5
 288     done
 289     die $LINENO "Time out while waiting for VM launch"
 290 }
 291
 292 inject_failure() {
 293     echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..."
 294     cat > disable_network.sh << 'END_TXT'
 295 #!/bin/bash -x
 296 dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $NF}')
 297 sleep 1
 298 sudo ip link set $dev down
 299 echo "doctor set link down at" $(date "+%s.%N")
 300 sleep 180
 301 sudo ip link set $dev up
 302 sleep 1
 303 END_TXT
 304     sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh
 305     chmod +x disable_network.sh
 306     scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:"
 307     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &'
 308     # use host time to get rid of potential time sync deviation between nodes
 309     triggered=$(date "+%s.%N")
 310 }
 311
 312 wait_consumer() {
 313     local interval=1
 314     local rounds=$(($1 / $interval))
 315     for i in `seq $rounds`; do
 316         notified_count=$(grep "doctor consumer notified at" consumer.log | wc -l)
 317         if [[ $notified_count -eq  $VMS_ON_FAILED_HOST ]]; then
 318             return 0
 319         fi
 320         sleep $interval
 321     done
 322     die $LINENO "Consumer hasn't received fault notification."
 323 }
 324
 325 calculate_notification_time() {
 326     wait_consumer 60
 327     #keep 'at' as the last keyword just before the value, and
 328     #use regex to get value instead of the fixed column
 329     detected=$(grep "doctor monitor detected at" monitor.log |\
 330                sed -e "s/^.* at //")
 331     notified=$(grep "doctor consumer notified at" consumer.log |\
 332                sed -e "s/^.* at //" | tail -1)
 333
 334     echo "$notified $detected" | \
 335         awk '{
 336             d = $1 - $2;
 337             if (d < 1 && d > 0) { print d " OK"; exit 0 }
 338             else { print d " NG"; exit 1 }
 339         }'
 340 }
 341
 342 check_host_status() {
 343     # Check host related to first Doctor VM is in wanted state
 344     # $1    Expected state
 345     # $2    Seconds to wait to have wanted state
 346     expected_state=$1
 347     local interval=5
 348     local rounds=$(($2 / $interval))
 349     for i in `seq $rounds`; do
 350         host_status_line=$(openstack $as_doctor_user --os-compute-api-version \
 351                            2.16 server show ${VM_BASENAME}1 | grep "host_status")
 352         host_status=$(echo $host_status_line | awk '{print $4}')
 353         die_if_not_set $LINENO host_status "host_status not reported by: nova show ${VM_BASENAME}1"
 354         if [[ "$expected_state" =~ "$host_status" ]] ; then
 355             echo "${VM_BASENAME}1 showing host_status: $host_status"
 356             return 0
 357         else
 358             sleep $interval
 359         fi
 360     done
 361     if [[ "$expected_state" =~ "$host_status" ]] ; then
 362         echo "${VM_BASENAME}1 showing host_status: $host_status"
 363     else
 364         die $LINENO  "host_status:$host_status not equal to expected_state: $expected_state"
 365     fi
 366 }
 367
 368 unset_forced_down_hosts() {
 369     # for debug
 370     openstack compute service list --service nova-compute
 371
 372     downed_computes=$(openstack compute service list --service nova-compute \
 373                       -f value -c Host -c State | grep ' down$' \
 374                       | sed -e 's/ *down$//')
 375     echo "downed_computes: $downed_computes"
 376     for host in $downed_computes
 377     do
 378         # TODO(r-mibu): use openstack client
 379         #openstack compute service set --up $host nova-compute
 380         nova service-force-down --unset $host nova-compute
 381     done
 382
 383     echo "waiting disabled compute host back to be enabled..."
 384     wait_until 'openstack compute service list --service nova-compute
 385                 -f value -c State | grep -q down' 240 5
 386
 387     for host in $downed_computes
 388     do
 389         # TODO(r-mibu): improve 'get_compute_ip_from_hostname'
 390         get_compute_ip_from_hostname $host
 391         wait_until "! ping -c 1 $COMPUTE_IP" 120 5
 392     done
 393 }
 394
 395 collect_logs() {
 396     if [[ -n "$COMPUTE_IP" ]];then
 397         scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
 398     fi
 399
 400     # TODO(yujunz) collect other logs, e.g. nova, aodh
 401 }
 402
 403 run_profiler() {
 404     if [[ "$PROFILER_TYPE" == "poc" ]]; then
 405         linkdown=$(grep "doctor set link down at " disable_network.log |\
 406                   sed -e "s/^.* at //")
 407         vmdown=$(grep "doctor mark vm.* error at" inspector.log |tail -n 1 |\
 408                  sed -e "s/^.* at //")
 409         hostdown=$(grep "doctor mark host.* down at" inspector.log |\
 410                  sed -e "s/^.* at //")
 411
 412         # TODO(yujunz) check the actual delay to verify time sync status
 413         # expected ~1s delay from $trigger to $linkdown
 414         relative_start=${linkdown}
 415         export DOCTOR_PROFILER_T00=$(python -c \
 416           "print(int(($linkdown-$relative_start)*1000))")
 417         export DOCTOR_PROFILER_T01=$(python -c \
 418           "print(int(($detected-$relative_start)*1000))")
 419         export DOCTOR_PROFILER_T03=$(python -c \
 420           "print(int(($vmdown-$relative_start)*1000))")
 421         export DOCTOR_PROFILER_T04=$(python -c \
 422           "print(int(($hostdown-$relative_start)*1000))")
 423         export DOCTOR_PROFILER_T09=$(python -c \
 424           "print(int(($notified-$relative_start)*1000))")
 425
 426         python profiler-poc.py >doctor_profiler.log 2>&1
 427     fi
 428 }
 429
 430 cleanup() {
 431     set +e
 432     echo "cleanup..."
 433     stop_monitor
 434     stop_inspector
 435     stop_consumer
 436
 437     unset_forced_down_hosts
 438     collect_logs
 439
 440     vms=$(openstack $as_doctor_user server list)
 441     vmstodel=""
 442     for i in `seq $VM_COUNT`; do
 443         $(echo "${vms}" | grep -q " $VM_BASENAME$i ") &&
 444         vmstodel+=" $VM_BASENAME$i"
 445     done
 446     [[ $vmstodel ]] && openstack $as_doctor_user server delete $vmstodel
 447     alarm_list=$($ceilometer alarm-list)
 448     for i in `seq $VM_COUNT`; do
 449         alarm_id=$(echo "${alarm_list}" | grep " $ALARM_BASENAME$i " |
 450                    awk '{print $2}')
 451         [ -n "$alarm_id" ] && $ceilometer alarm-delete "$alarm_id"
 452     done
 453     openstack $as_doctor_user subnet delete $NET_NAME
 454     sleep 1
 455     openstack $as_doctor_user network delete $NET_NAME
 456     sleep 1
 457
 458     image_id=$(openstack image list | grep " $IMAGE_NAME " | awk '{print $2}')
 459     sleep 1
 460     #if an existing image was used, there's no need to remove it here
 461     if [[ "$use_existing_image" == false ]] ; then
 462         [ -n "$image_id" ] && openstack image delete "$image_id"
 463     fi
 464
 465     remove_test_user
 466
 467     cleanup_installer
 468     cleanup_inspector
 469
 470     # NOTE: Temporal log printer.
 471     for f in $(find . -name '*.log')
 472     do
 473         echo
 474         echo "[$f]"
 475         sed -e 's/^/ | /' $f
 476         echo
 477     done
 478 }
 479
 480 # Main process
 481
 482 echo "Note: doctor/tests/run.sh has been executed."
 483 git log --oneline -1 || true   # ignore even you don't have git installed
 484
 485 trap cleanup EXIT
 486
 487 source $TOP_DIR/functions-common
 488 source $TOP_DIR/lib/installer
 489 source $TOP_DIR/lib/inspector
 490
 491 setup_installer
 492
 493 echo "preparing VM image..."
 494 download_image
 495 register_image
 496
 497 echo "creating test user..."
 498 create_test_user
 499
 500 echo "creating VM..."
 501 boot_vm
 502 wait_for_vm_launch
 503
 504 echo "creating alarm..."
 505 #TODO: change back to use, network problems depends on infra and installers
 506 #get_consumer_ip
 507 create_alarm
 508
 509 echo "starting doctor sample components..."
 510 start_inspector
 511 start_monitor
 512 start_consumer
 513
 514 sleep 60
 515 echo "injecting host failure..."
 516 inject_failure
 517
 518 check_host_status "(DOWN|UNKNOWN)" 60
 519 calculate_notification_time
 520 unset_forced_down_hosts
 521 collect_logs
 522 run_profiler
 523
 524 echo "done"