2 ##############################################################################
3 # Copyright (c) 2016 NEC Corporation and others.
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 ##############################################################################
13 [[ "${CI_DEBUG:-true}" == [Tt]rue ]] && set -x
15 IMAGE_URL=https://launchpad.net/cirros/trunk/0.3.0/+download/cirros-0.3.0-x86_64-disk.img
16 #if an existing image name is provided in the enviroment, use that one
17 IMAGE_NAME=${IMAGE_NAME:-cirros}
18 IMAGE_FILE="${IMAGE_NAME}.img"
22 ALARM_NAME=doctor_alarm1
28 #TODO: change back to `_member_` when JIRA DOCTOR-55 is done
30 PROFILER_TYPE=${PROFILER_TYPE:-none}
32 TOP_DIR=$(cd $(dirname "$0") && pwd)
34 ssh_opts="-o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no"
35 as_doctor_user="--os-username $DOCTOR_USER --os-password $DOCTOR_PW
36 --os-tenant-name $DOCTOR_PROJECT"
41 get_compute_host_info() {
42 # get computer host info which VM boot in
43 COMPUTE_HOST=$(openstack $as_doctor_user server show $VM_NAME |
44 grep "OS-EXT-SRV-ATTR:host" | awk '{ print $4 }')
45 compute_host_in_undercloud=${COMPUTE_HOST%%.*}
46 die_if_not_set $LINENO COMPUTE_HOST "Failed to get compute hostname"
48 if is_installer apex; then
49 COMPUTE_USER=${COMPUTE_USER:-heat-admin}
50 COMPUTE_IP=$(sudo ssh $ssh_opts $INSTALLER_IP \
52 nova show $compute_host_in_undercloud \
53 | awk '/ ctlplane network /{print \$5}'")
54 elif is_installer fuel; then
55 COMPUTE_USER=${COMPUTE_USER:-root}
56 node_id=$(echo $compute_host_in_undercloud | cut -d "-" -f 2)
57 COMPUTE_IP=$(sshpass -p r00tme ssh 2>/dev/null $ssh_opts root@${INSTALLER_IP} \
58 "fuel node|awk -F '|' -v id=$node_id '{if (\$1 == id) print \$5}' |xargs")
59 elif is_installer local; then
60 COMPUTE_USER=${COMPUTE_USER:-$(whoami)}
61 COMPUTE_IP=$(getent hosts "$COMPUTE_HOST" | awk '{ print $1 }')
64 die_if_not_set $LINENO COMPUTE_IP "Could not resolve $COMPUTE_HOST. Either manually set COMPUTE_IP or enable DNS resolution."
65 echo "COMPUTE_HOST=$COMPUTE_HOST"
66 echo "COMPUTE_IP=$COMPUTE_IP"
68 # verify connectivity to target compute host
69 ping -c 1 "$COMPUTE_IP"
70 if [[ $? -ne 0 ]] ; then
71 die $LINENO "Can not ping to computer host"
74 # verify ssh to target compute host
75 ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'exit'
76 if [[ $? -ne 0 ]] ; then
77 die $LINENO "Can not ssh to computer host"
82 local get_consumer_command="ip route get $COMPUTE_IP | awk '/ src /{print \$NF}'"
83 if is_installer apex; then
84 CONSUMER_IP=$(sudo ssh $ssh_opts root@$INSTALLER_IP \
85 "$get_consumer_command")
86 elif is_installer fuel; then
87 CONSUMER_IP=$(sudo sshpass -p r00tme ssh $ssh_opts root@${INSTALLER_IP} \
88 "$get_consumer_command")
89 elif is_installer local; then
90 CONSUMER_IP=`$get_consumer_command`
92 echo "CONSUMER_IP=$CONSUMER_IP"
94 die_if_not_set $LINENO CONSUMER_IP "Could not get CONSUMER_IP."
98 #if a different name was provided for the image in the enviroment there's no need to download the image
99 use_existing_image=false
100 openstack image list | grep -q " $IMAGE_NAME " && use_existing_image=true
102 if [[ "$use_existing_image" == false ]] ; then
103 [ -e "$IMAGE_FILE" ] && return 0
104 wget "$IMAGE_URL" -o "$IMAGE_FILE"
109 openstack image list | grep -q " $IMAGE_NAME " && return 0
110 openstack image create "$IMAGE_NAME" \
112 --disk-format "$IMAGE_FORMAT" \
113 --container-format bare \
118 openstack project list | grep -q " $DOCTOR_PROJECT " || {
119 openstack project create "$DOCTOR_PROJECT"
121 openstack user list | grep -q " $DOCTOR_USER " || {
122 openstack user create "$DOCTOR_USER" --password "$DOCTOR_PW" \
123 --project "$DOCTOR_PROJECT"
125 openstack user role list "$DOCTOR_USER" --project "$DOCTOR_PROJECT" \
126 | grep -q " $DOCTOR_ROLE " || {
127 openstack role add "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
128 --project "$DOCTOR_PROJECT"
133 # test VM done with test user, so can test non-admin
134 openstack $as_doctor_user server list | grep -q " $VM_NAME " && return 0
135 openstack $as_doctor_user server create --flavor "$VM_FLAVOR" \
136 --image "$IMAGE_NAME" \
142 # get vm_id as test user
143 ceilometer $as_doctor_user alarm-list | grep -q " $ALARM_NAME " && return 0
144 vm_id=$(openstack $as_doctor_user server list | grep " $VM_NAME " | awk '{print $2}')
145 # TODO(r-mibu): change notification endpoint from localhost to the consumer
146 # IP address (functest container).
147 ceilometer $as_doctor_user alarm-event-create --name "$ALARM_NAME" \
148 --alarm-action "http://localhost:$CONSUMER_PORT/failure" \
149 --description "VM failure" \
151 --repeat-actions False \
152 --severity "moderate" \
153 --event-type compute.instance.update \
154 -q "traits.state=string::error; traits.instance_id=string::$vm_id"
158 pgrep -f "python monitor.py" && return 0
159 sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
160 "http://127.0.0.1:$INSPECTOR_PORT/events" > monitor.log 2>&1 &
164 pgrep -f "python monitor.py" || return 0
165 sudo kill $(pgrep -f "python monitor.py")
169 pgrep -f "python consumer.py" && return 0
170 python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 &
172 # NOTE(r-mibu): create tunnel to the controller nodes, so that we can
173 # avoid some network problems dpends on infra and installers.
174 # This tunnel will be terminated by stop_consumer() or after 10 mins passed.
175 if ! is_installer local; then
176 if is_installer apex; then
177 CONTROLLER_IPS=$(sudo ssh $ssh_opts $INSTALLER_IP \
179 nova list | grep ' overcloud-controller-[0-9] ' \
180 | sed -e 's/^.*ctlplane=//' -e 's/ *|\$//'")
181 elif is_installer fuel; then
185 die_if_not_set $LINENO CONTROLLER_IPS "Could not get CONTROLLER_IPS."
186 for ip in $CONTROLLER_IPS
188 forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
189 tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
190 $tunnel_command > "ssh_tunnel.${ip}.log" 2>&1 < /dev/null &
196 pgrep -f "python consumer.py" || return 0
197 kill $(pgrep -f "python consumer.py")
199 # NOTE(r-mibu): terminate tunnels to the controller nodes
200 if ! is_installer local; then
201 for ip in $CONTROLLER_IPS
203 forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
204 tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
205 kill $(pgrep -f "$tunnel_command")
210 wait_for_vm_launch() {
211 echo "waiting for vm launch..."
214 while [[ ${count} -lt 60 ]]
216 state=$(openstack $as_doctor_user server list | grep " $VM_NAME " | awk '{print $6}')
217 if [[ "$state" == "ACTIVE" ]]; then
218 # NOTE(cgoncalves): sleeping for a bit to stabilize
219 # See python-openstackclient/functional/tests/compute/v2/test_server.py:wait_for_status
223 if [[ "$state" == "ERROR" ]]; then
224 openstack $as_doctor_user server show $VM_NAME
225 die $LINENO "vm state is ERROR"
230 die $LINENO "Time out while waiting for VM launch"
234 echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..."
235 cat > disable_network.sh << 'END_TXT'
237 dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $7}')
238 [[ -n "$dev" ]] || dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $5}')
240 sudo ip link set $dev down
241 echo "doctor set host down at" $(date "+%s.%N")
243 sudo ip link set $dev up
246 sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh
247 chmod +x disable_network.sh
248 scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:"
249 ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &'
252 profile_performance_poc() {
253 triggered=$(grep "^doctor set host down at" disable_network.log |\
254 sed -e "s/^.* at //")
255 vmdown=$(grep "doctor mark vm.* error at" inspector.log |tail -n 1 |\
256 sed -e "s/^.* at //")
257 hostdown=$(grep "doctor mark host.* down at" inspector.log |\
258 sed -e "s/^.* at //")
260 #calculate the relative interval to triggered(T00)
261 export DOCTOR_PROFILER_T00=0
262 export DOCTOR_PROFILER_T01=$(echo "($detected-$triggered)*1000/1" |bc)
263 export DOCTOR_PROFILER_T03=$(echo "($vmdown-$triggered)*1000/1" |bc)
264 export DOCTOR_PROFILER_T04=$(echo "($hostdown-$triggered)*1000/1" |bc)
265 export DOCTOR_PROFILER_T09=$(echo "($notified-$triggered)*1000/1" |bc)
267 python profiler-poc.py
270 calculate_notification_time() {
271 if ! grep -q "doctor consumer notified at" consumer.log ; then
272 die $LINENO "Consumer hasn't received fault notification."
275 #keep 'at' as the last keyword just before the value, and
276 #use regex to get value instead of the fixed column
277 detected=$(grep "doctor monitor detected at" monitor.log |\
278 sed -e "s/^.* at //")
279 notified=$(grep "doctor consumer notified at" consumer.log |\
280 sed -e "s/^.* at //")
282 if [[ "$PROFILER_TYPE" == "poc" ]]; then
283 profile_performance_poc
286 echo "$notified $detected" | \
289 if (d < 1 && d > 0) { print d " OK"; exit 0 }
290 else { print d " NG"; exit 1 }
294 check_host_status() {
297 host_status_line=$(openstack $as_doctor_user --os-compute-api-version 2.16 \
298 server show $VM_NAME | grep "host_status")
299 host_status=$(echo $host_status_line | awk '{print $4}')
300 die_if_not_set $LINENO host_status "host_status not reported by: nova show $VM_NAME"
301 if [[ "$expected_state" =~ "$host_status" ]] ; then
302 echo "$VM_NAME showing host_status: $host_status"
304 die $LINENO "host_status:$host_status not equal to expected_state: $expected_state"
308 unset_forced_down_hosts() {
309 for host in $(openstack compute service list --service nova-compute \
310 -f value -c Host -c State | sed -n -e '/down$/s/ *down$//p')
312 # TODO (r-mibu): make sample inspector use keystone v3 api
313 OS_AUTH_URL=${OS_AUTH_URL/v3/v2.0} \
314 python ./nova_force_down.py $host --unset
317 echo "waiting disabled compute host back to be enabled..."
318 wait_until 'openstack compute service list --service nova-compute
319 -f value -c State | grep -q down' 240 5
329 unset_forced_down_hosts
330 # TODO: We need to make sure the target compute host is back to IP
331 # reachable. wait_ping() will be added by tojuvone .
333 scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
335 openstack $as_doctor_user server list | grep -q " $VM_NAME " && openstack $as_doctor_user server delete "$VM_NAME"
337 alarm_id=$(ceilometer $as_doctor_user alarm-list | grep " $ALARM_NAME " | awk '{print $2}')
339 [ -n "$alarm_id" ] && ceilometer $as_doctor_user alarm-delete "$alarm_id"
342 image_id=$(openstack image list | grep " $IMAGE_NAME " | awk '{print $2}')
344 #if an existing image was used, there's no need to remove it here
345 if [[ "$use_existing_image" == false ]] ; then
346 [ -n "$image_id" ] && openstack image delete "$image_id"
348 openstack role remove "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
349 --project "$DOCTOR_PROJECT"
350 openstack project delete "$DOCTOR_PROJECT"
351 openstack user delete "$DOCTOR_USER"
359 echo "Note: doctor/tests/run.sh has been executed."
360 git log --oneline -1 || true # ignore even you don't have git installed
364 source $TOP_DIR/functions-common
365 source $TOP_DIR/lib/installer
366 source $TOP_DIR/lib/inspector
370 echo "preparing VM image..."
374 echo "creating test user..."
377 echo "creating VM..."
381 echo "get computer host info..."
382 get_compute_host_info
384 echo "creating alarm..."
385 #TODO: change back to use, network problems depends on infra and installers
389 echo "starting doctor sample components..."
395 echo "injecting host failure..."
399 check_host_status "(DOWN|UNKNOWN)"
400 calculate_notification_time