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 as_doctor_user="--os-username $DOCTOR_USER --os-password $DOCTOR_PW
  35                 --os-tenant-name $DOCTOR_PROJECT"
  36
  37
  38 # Functions
  39
  40 get_compute_host_info() {
  41     # get computer host info which VM boot in
  42     COMPUTE_HOST=$(openstack $as_doctor_user server show $VM_NAME |
  43                    grep "OS-EXT-SRV-ATTR:host" | awk '{ print $4 }')
  44     compute_host_in_undercloud=${COMPUTE_HOST%%.*}
  45     die_if_not_set $LINENO COMPUTE_HOST "Failed to get compute hostname"
  46
  47     get_compute_ip_from_hostname $COMPUTE_HOST
  48
  49     echo "COMPUTE_HOST=$COMPUTE_HOST"
  50     echo "COMPUTE_IP=$COMPUTE_IP"
  51
  52     # verify connectivity to target compute host
  53     ping -c 1 "$COMPUTE_IP"
  54     if [[ $? -ne 0 ]] ; then
  55         die $LINENO "Can not ping to computer host"
  56     fi
  57
  58     # verify ssh to target compute host
  59     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'exit'
  60     if [[ $? -ne 0 ]] ; then
  61         die $LINENO "Can not ssh to computer host"
  62     fi
  63 }
  64
  65 # TODO(r-mibu): update this function to support consumer instance
  66 #               and migrate this function into installer lib
  67 get_consumer_ip___to_be_removed() {
  68     local get_consumer_command="ip route get $COMPUTE_IP | awk '/ src /{print \$NF}'"
  69     if is_installer apex; then
  70         CONSUMER_IP=$(sudo ssh $ssh_opts root@$INSTALLER_IP \
  71                       "$get_consumer_command")
  72     elif is_installer fuel; then
  73         CONSUMER_IP=$(sudo sshpass -p r00tme ssh $ssh_opts root@${INSTALLER_IP} \
  74                       "$get_consumer_command")
  75     elif is_installer local; then
  76         CONSUMER_IP=`$get_consumer_command`
  77     fi
  78     echo "CONSUMER_IP=$CONSUMER_IP"
  79
  80     die_if_not_set $LINENO CONSUMER_IP "Could not get CONSUMER_IP."
  81 }
  82
  83 download_image() {
  84     #if a different name was provided for the image in the enviroment there's no need to download the image
  85     use_existing_image=false
  86     openstack image list | grep -q " $IMAGE_NAME " && use_existing_image=true
  87
  88     if [[ "$use_existing_image" == false ]] ; then
  89         [ -e "$IMAGE_FILE" ] && return 0
  90         wget "$IMAGE_URL" -o "$IMAGE_FILE"
  91     fi
  92 }
  93
  94 register_image() {
  95     openstack image list | grep -q " $IMAGE_NAME " && return 0
  96     openstack image create "$IMAGE_NAME" \
  97                            --public \
  98                            --disk-format "$IMAGE_FORMAT" \
  99                            --container-format bare \
 100                            --file "$IMAGE_FILE"
 101 }
 102
 103 create_test_user() {
 104     openstack project list | grep -q " $DOCTOR_PROJECT " || {
 105         openstack project create "$DOCTOR_PROJECT"
 106     }
 107     openstack user list | grep -q " $DOCTOR_USER " || {
 108         openstack user create "$DOCTOR_USER" --password "$DOCTOR_PW" \
 109                               --project "$DOCTOR_PROJECT"
 110     }
 111     openstack user role list "$DOCTOR_USER" --project "$DOCTOR_PROJECT" \
 112     | grep -q " $DOCTOR_ROLE " || {
 113         openstack role add "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 114                            --project "$DOCTOR_PROJECT"
 115     }
 116 }
 117
 118 boot_vm() {
 119     # test VM done with test user, so can test non-admin
 120     openstack $as_doctor_user server list | grep -q " $VM_NAME " && return 0
 121     openstack $as_doctor_user server create --flavor "$VM_FLAVOR" \
 122                             --image "$IMAGE_NAME" \
 123                             "$VM_NAME"
 124     sleep 1
 125 }
 126
 127 create_alarm() {
 128     # get vm_id as test user
 129     ceilometer $as_doctor_user alarm-list | grep -q " $ALARM_NAME " && return 0
 130     vm_id=$(openstack $as_doctor_user server list | grep " $VM_NAME " | awk '{print $2}')
 131     # TODO(r-mibu): change notification endpoint from localhost to the consumer
 132     # IP address (functest container).
 133     ceilometer $as_doctor_user alarm-event-create --name "$ALARM_NAME" \
 134         --alarm-action "http://localhost:$CONSUMER_PORT/failure" \
 135         --description "VM failure" \
 136         --enabled True \
 137         --repeat-actions False \
 138         --severity "moderate" \
 139         --event-type compute.instance.update \
 140         -q "traits.state=string::error; traits.instance_id=string::$vm_id"
 141 }
 142
 143 start_monitor() {
 144     pgrep -f "python monitor.py" && return 0
 145     sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
 146         "http://127.0.0.1:$INSPECTOR_PORT/events" > monitor.log 2>&1 &
 147 }
 148
 149 stop_monitor() {
 150     pgrep -f "python monitor.py" || return 0
 151     sudo kill $(pgrep -f "python monitor.py")
 152 }
 153
 154 start_consumer() {
 155     pgrep -f "python consumer.py" && return 0
 156     python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 &
 157
 158     # NOTE(r-mibu): create tunnel to the controller nodes, so that we can
 159     # avoid some network problems dpends on infra and installers.
 160     # This tunnel will be terminated by stop_consumer() or after 10 mins passed.
 161     if ! is_installer local; then
 162         for ip in $CONTROLLER_IPS
 163         do
 164             forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
 165             tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
 166             $tunnel_command > "ssh_tunnel.${ip}.log" 2>&1 < /dev/null &
 167         done
 168     fi
 169 }
 170
 171 stop_consumer() {
 172     pgrep -f "python consumer.py" || return 0
 173     kill $(pgrep -f "python consumer.py")
 174
 175     # NOTE(r-mibu): terminate tunnels to the controller nodes
 176     if ! is_installer local; then
 177         for ip in $CONTROLLER_IPS
 178         do
 179             forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
 180             tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
 181             kill $(pgrep -f "$tunnel_command")
 182         done
 183     fi
 184 }
 185
 186 wait_for_vm_launch() {
 187     echo "waiting for vm launch..."
 188
 189     count=0
 190     while [[ ${count} -lt 60 ]]
 191     do
 192         state=$(openstack $as_doctor_user server list | grep " $VM_NAME " | awk '{print $6}')
 193         if [[ "$state" == "ACTIVE" ]]; then
 194             # NOTE(cgoncalves): sleeping for a bit to stabilize
 195             # See python-openstackclient/functional/tests/compute/v2/test_server.py:wait_for_status
 196             sleep 5
 197             return 0
 198         fi
 199         if [[ "$state" == "ERROR" ]]; then
 200             openstack $as_doctor_user server show $VM_NAME
 201             die $LINENO "vm state is ERROR"
 202         fi
 203         count=$(($count+1))
 204         sleep 1
 205     done
 206     die $LINENO "Time out while waiting for VM launch"
 207 }
 208
 209 inject_failure() {
 210     echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..."
 211     cat > disable_network.sh << 'END_TXT'
 212 #!/bin/bash -x
 213 dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $7}')
 214 [[ -n "$dev" ]] || dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $5}')
 215 sleep 1
 216 sudo ip link set $dev down
 217 echo "doctor set host down at" $(date "+%s.%N")
 218 sleep 180
 219 sudo ip link set $dev up
 220 sleep 1
 221 END_TXT
 222     sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh
 223     chmod +x disable_network.sh
 224     scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:"
 225     ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &'
 226 }
 227
 228 profile_performance_poc() {
 229     triggered=$(grep "^doctor set host down at" disable_network.log |\
 230                 sed -e "s/^.* at //")
 231     vmdown=$(grep "doctor mark vm.* error at" inspector.log |tail -n 1 |\
 232                sed -e "s/^.* at //")
 233     hostdown=$(grep "doctor mark host.* down at" inspector.log |\
 234                sed -e "s/^.* at //")
 235
 236     #calculate the relative interval to triggered(T00)
 237     export DOCTOR_PROFILER_T00=0
 238     export DOCTOR_PROFILER_T01=$(echo "($detected-$triggered)*1000/1" |bc)
 239     export DOCTOR_PROFILER_T03=$(echo "($vmdown-$triggered)*1000/1" |bc)
 240     export DOCTOR_PROFILER_T04=$(echo "($hostdown-$triggered)*1000/1" |bc)
 241     export DOCTOR_PROFILER_T09=$(echo "($notified-$triggered)*1000/1" |bc)
 242
 243     python profiler-poc.py
 244 }
 245
 246 calculate_notification_time() {
 247     if ! grep -q "doctor consumer notified at" consumer.log ; then
 248         die $LINENO "Consumer hasn't received fault notification."
 249     fi
 250
 251     #keep 'at' as the last keyword just before the value, and
 252     #use regex to get value instead of the fixed column
 253     detected=$(grep "doctor monitor detected at" monitor.log |\
 254                sed -e "s/^.* at //")
 255     notified=$(grep "doctor consumer notified at" consumer.log |\
 256                sed -e "s/^.* at //")
 257
 258     if [[ "$PROFILER_TYPE" == "poc" ]]; then
 259         profile_performance_poc
 260     fi
 261
 262     echo "$notified $detected" | \
 263         awk '{
 264             d = $1 - $2;
 265             if (d < 1 && d > 0) { print d " OK"; exit 0 }
 266             else { print d " NG"; exit 1 }
 267         }'
 268 }
 269
 270 check_host_status() {
 271     expected_state=$1
 272
 273     host_status_line=$(openstack $as_doctor_user --os-compute-api-version 2.16 \
 274                        server show $VM_NAME | grep "host_status")
 275     host_status=$(echo $host_status_line | awk '{print $4}')
 276     die_if_not_set $LINENO host_status "host_status not reported by: nova show $VM_NAME"
 277     if [[ "$expected_state" =~ "$host_status" ]] ; then
 278         echo "$VM_NAME showing host_status: $host_status"
 279     else
 280         die $LINENO "host_status:$host_status not equal to expected_state: $expected_state"
 281     fi
 282 }
 283
 284 unset_forced_down_hosts() {
 285     for host in $(openstack compute service list --service nova-compute \
 286                   -f value -c Host -c State | sed -n -e '/down$/s/ *down$//p')
 287     do
 288         # TODO (r-mibu): make sample inspector use keystone v3 api
 289         OS_AUTH_URL=${OS_AUTH_URL/v3/v2.0} \
 290         python ./nova_force_down.py $host --unset
 291     done
 292
 293     echo "waiting disabled compute host back to be enabled..."
 294     wait_until 'openstack compute service list --service nova-compute
 295                 -f value -c State | grep -q down' 240 5
 296 }
 297
 298 cleanup() {
 299     set +e
 300     echo "cleanup..."
 301     stop_monitor
 302     stop_inspector
 303     stop_consumer
 304
 305     unset_forced_down_hosts
 306     # TODO: We need to make sure the target compute host is back to IP
 307     #       reachable. wait_ping() will be added by tojuvone .
 308     sleep 110
 309     if is_set COMPUTE_IP; then
 310         scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
 311     fi
 312
 313     openstack $as_doctor_user server list | grep -q " $VM_NAME " && openstack $as_doctor_user server delete "$VM_NAME"
 314     sleep 1
 315     alarm_id=$(ceilometer $as_doctor_user alarm-list | grep " $ALARM_NAME " | awk '{print $2}')
 316     sleep 1
 317     [ -n "$alarm_id" ] && ceilometer $as_doctor_user alarm-delete "$alarm_id"
 318     sleep 1
 319
 320     image_id=$(openstack image list | grep " $IMAGE_NAME " | awk '{print $2}')
 321     sleep 1
 322     #if an existing image was used, there's no need to remove it here
 323     if [[ "$use_existing_image" == false ]] ; then
 324         [ -n "$image_id" ] && openstack image delete "$image_id"
 325     fi
 326     openstack role remove "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
 327                               --project "$DOCTOR_PROJECT"
 328     openstack project delete "$DOCTOR_PROJECT"
 329     openstack user delete "$DOCTOR_USER"
 330
 331     cleanup_installer
 332     cleanup_inspector
 333 }
 334
 335 # Main process
 336
 337 echo "Note: doctor/tests/run.sh has been executed."
 338 git log --oneline -1 || true   # ignore even you don't have git installed
 339
 340 trap cleanup EXIT
 341
 342 source $TOP_DIR/functions-common
 343 source $TOP_DIR/lib/installer
 344 source $TOP_DIR/lib/inspector
 345
 346 setup_installer
 347
 348 echo "preparing VM image..."
 349 download_image
 350 register_image
 351
 352 echo "creating test user..."
 353 create_test_user
 354
 355 echo "creating VM..."
 356 boot_vm
 357 wait_for_vm_launch
 358
 359 echo "get computer host info..."
 360 get_compute_host_info
 361
 362 echo "creating alarm..."
 363 #TODO: change back to use, network problems depends on infra and installers
 364 #get_consumer_ip
 365 create_alarm
 366
 367 echo "starting doctor sample components..."
 368 start_inspector
 369 start_monitor
 370 start_consumer
 371
 372 sleep 60
 373 echo "injecting host failure..."
 374 inject_failure
 375 sleep 60
 376
 377 check_host_status "(DOWN|UNKNOWN)"
 378 calculate_notification_time
 379
 380 echo "done"