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