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="${IMAGE_NAME}.img"
22 #if VM_COUNT set, use that instead
23 VM_COUNT=${VM_COUNT:-1}
25 NET_CIDR=192.168.168.0/24
26 ALARM_BASENAME=doctor_alarm
32 PROFILER_TYPE=${PROFILER_TYPE:-none}
33 PYTHON_ENABLE=${PYTHON_ENABLE:-false}
35 TOP_DIR=$(cd $(dirname "$0") && pwd)
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 $DOCTOR_PROJECT"
48 get_compute_host_info() {
49 # get computer host info which first VM boot in as admin user
50 COMPUTE_HOST=$(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"
55 get_compute_ip_from_hostname $COMPUTE_HOST
57 echo "COMPUTE_HOST=$COMPUTE_HOST"
58 echo "COMPUTE_IP=$COMPUTE_IP"
60 # verify connectivity to target compute host
61 ping -c 1 "$COMPUTE_IP"
62 if [[ $? -ne 0 ]] ; then
63 die $LINENO "Can not ping to computer host"
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"
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`
86 echo "CONSUMER_IP=$CONSUMER_IP"
88 die_if_not_set $LINENO CONSUMER_IP "Could not get CONSUMER_IP."
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 " $IMAGE_NAME " && use_existing_image=true
96 if [[ "$use_existing_image" == false ]] ; then
97 [ -e "$IMAGE_FILE" ] && return 0
98 wget "$IMAGE_URL" -o "$IMAGE_FILE"
103 openstack image list | grep -q " $IMAGE_NAME " && return 0
104 openstack image create "$IMAGE_NAME" \
106 --disk-format "$IMAGE_FORMAT" \
107 --container-format bare \
112 openstack project list | grep -q " $DOCTOR_PROJECT " || {
113 openstack project create --description "Doctor Project" \
116 openstack user list | grep -q " $DOCTOR_USER " || {
117 openstack user create "$DOCTOR_USER" --password "$DOCTOR_PW" \
118 --project "$DOCTOR_PROJECT"
120 openstack role show "$DOCTOR_ROLE" | grep -q " $DOCTOR_ROLE " || {
121 openstack role create "$DOCTOR_ROLE"
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"
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"
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
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 " | \
141 if [ $OLD_INSTANCE_QUOTA -lt $VM_COUNT ]; then
142 openstack quota set --instances $VM_COUNT \
145 OLD_CORES_QUOTA=$(echo "${DOCTOR_QUOTA}" | grep " cores " | \
147 if [ $OLD_CORES_QUOTA -lt $VM_COUNT ]; then
148 openstack quota set --cores $VM_COUNT \
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"
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"
165 openstack user delete "$DOCTOR_USER"
167 openstack project delete "$DOCTOR_PROJECT"
172 # test VM done with test user, so can test non-admin
174 if ! openstack $as_doctor_user network show $NET_NAME; then
175 openstack $as_doctor_user network create $NET_NAME
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 $NET_CIDR --no-dhcp
181 net_id=$(openstack $as_doctor_user network show $NET_NAME -f value -c id)
183 servers=$(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"
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" \
206 --repeat-actions False \
207 --severity "moderate" \
208 --event-type compute.instance.update \
209 -q "traits.state=string::error; \
210 traits.instance_id=string::$vm_id"
216 pgrep -f "python monitor.py" && return 0
217 sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
222 pgrep -f "python monitor.py" || return 0
223 sudo kill $(pgrep -f "python monitor.py")
227 pgrep -f "python consumer.py" && return 0
228 python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 &
230 # NOTE(r-mibu): create tunnel to the controller nodes, so that we can
231 # avoid some network problems dpends on infra and installers.
232 # This tunnel will be terminated by stop_consumer() or after 10 mins passed.
233 if ! is_installer local; then
234 for ip in $CONTROLLER_IPS
236 forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
237 tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
238 $tunnel_command > "ssh_tunnel.${ip}.log" 2>&1 < /dev/null &
244 pgrep -f "python consumer.py" || return 0
245 kill $(pgrep -f "python consumer.py")
247 # NOTE(r-mibu): terminate tunnels to the controller nodes
248 if ! is_installer local; then
249 for ip in $CONTROLLER_IPS
251 forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
252 tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
253 kill $(pgrep -f "$tunnel_command")
258 wait_for_vm_launch() {
259 echo "waiting for vm launch..."
262 while [[ ${count} -lt 60 ]]
265 vms=$(openstack $as_doctor_user server list)
266 for i in `seq $VM_COUNT`; do
267 state=$(echo "${vms}" | grep " $VM_BASENAME$i " | awk '{print $6}')
268 if [[ "$state" == "ACTIVE" ]]; then
269 active_count=$(($active_count+1))
270 elif [[ "$state" == "ERROR" ]]; then
271 die $LINENO "vm state $VM_BASENAME$i is ERROR"
273 #This VM not yet active
279 [[ $active_count -eq $VM_COUNT ]] && {
280 echo "get computer host info..."
281 get_compute_host_info
282 VMS_ON_FAILED_HOST=$(openstack $as_doctor_user server list --host \
283 $COMPUTE_HOST | grep " ${VM_BASENAME}" | wc -l)
290 die $LINENO "Time out while waiting for VM launch"
294 echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..."
295 cat > disable_network.sh << 'END_TXT'
297 dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $NF}')
299 sudo ip link set $dev down
300 echo "doctor set link down at" $(date "+%s.%N")
302 sudo ip link set $dev up
305 sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh
306 chmod +x disable_network.sh
307 scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:"
308 ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &'
309 # use host time to get rid of potential time sync deviation between nodes
310 triggered=$(date "+%s.%N")
315 local rounds=$(($1 / $interval))
316 for i in `seq $rounds`; do
317 notified_count=$(grep "doctor consumer notified at" consumer.log | wc -l)
318 if [[ $notified_count -eq $VMS_ON_FAILED_HOST ]]; then
323 die $LINENO "Consumer hasn't received fault notification."
326 calculate_notification_time() {
328 #keep 'at' as the last keyword just before the value, and
329 #use regex to get value instead of the fixed column
330 detected=$(grep "doctor monitor detected at" monitor.log |\
331 sed -e "s/^.* at //")
332 notified=$(grep "doctor consumer notified at" consumer.log |\
333 sed -e "s/^.* at //" | tail -1)
335 echo "$notified $detected" | \
338 if (d < 1 && d > 0) { print d " OK"; exit 0 }
339 else { print d " NG"; exit 1 }
343 check_host_status() {
344 # Check host related to first Doctor VM is in wanted state
346 # $2 Seconds to wait to have wanted state
349 local rounds=$(($2 / $interval))
350 for i in `seq $rounds`; do
351 host_status_line=$(openstack $as_doctor_user --os-compute-api-version \
352 2.16 server show ${VM_BASENAME}1 | grep "host_status")
353 host_status=$(echo $host_status_line | awk '{print $4}')
354 die_if_not_set $LINENO host_status "host_status not reported by: nova show ${VM_BASENAME}1"
355 if [[ "$expected_state" =~ "$host_status" ]] ; then
356 echo "${VM_BASENAME}1 showing host_status: $host_status"
362 if [[ "$expected_state" =~ "$host_status" ]] ; then
363 echo "${VM_BASENAME}1 showing host_status: $host_status"
365 die $LINENO "host_status:$host_status not equal to expected_state: $expected_state"
369 unset_forced_down_hosts() {
371 openstack compute service list --service nova-compute
373 downed_computes=$(openstack compute service list --service nova-compute \
374 -f value -c Host -c State | grep ' down$' \
375 | sed -e 's/ *down$//')
376 echo "downed_computes: $downed_computes"
377 for host in $downed_computes
379 # TODO(r-mibu): use openstack client
380 #openstack compute service set --up $host nova-compute
381 nova service-force-down --unset $host nova-compute
384 echo "waiting disabled compute host back to be enabled..."
385 wait_until 'openstack compute service list --service nova-compute
386 -f value -c State | grep -q down' 240 5
388 for host in $downed_computes
390 # TODO(r-mibu): improve 'get_compute_ip_from_hostname'
391 get_compute_ip_from_hostname $host
392 wait_until "! ping -c 1 $COMPUTE_IP" 120 5
397 if [[ -n "$COMPUTE_IP" ]];then
398 scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
401 # TODO(yujunz) collect other logs, e.g. nova, aodh
405 if [[ "$PROFILER_TYPE" == "poc" ]]; then
406 linkdown=$(grep "doctor set link down at " disable_network.log |\
407 sed -e "s/^.* at //")
408 vmdown=$(grep "doctor mark vm.* error at" inspector.log |tail -n 1 |\
409 sed -e "s/^.* at //")
410 hostdown=$(grep "doctor mark host.* down at" inspector.log |\
411 sed -e "s/^.* at //")
413 # TODO(yujunz) check the actual delay to verify time sync status
414 # expected ~1s delay from $trigger to $linkdown
415 relative_start=${linkdown}
416 export DOCTOR_PROFILER_T00=$(python -c \
417 "print(int(($linkdown-$relative_start)*1000))")
418 export DOCTOR_PROFILER_T01=$(python -c \
419 "print(int(($detected-$relative_start)*1000))")
420 export DOCTOR_PROFILER_T03=$(python -c \
421 "print(int(($vmdown-$relative_start)*1000))")
422 export DOCTOR_PROFILER_T04=$(python -c \
423 "print(int(($hostdown-$relative_start)*1000))")
424 export DOCTOR_PROFILER_T09=$(python -c \
425 "print(int(($notified-$relative_start)*1000))")
427 python profiler-poc.py >doctor_profiler.log 2>&1
438 unset_forced_down_hosts
441 vms=$(openstack $as_doctor_user server list)
443 for i in `seq $VM_COUNT`; do
444 $(echo "${vms}" | grep -q " $VM_BASENAME$i ") &&
445 vmstodel+=" $VM_BASENAME$i"
447 [[ $vmstodel ]] && openstack $as_doctor_user server delete $vmstodel
448 alarm_list=$($ceilometer alarm-list)
449 for i in `seq $VM_COUNT`; do
450 alarm_id=$(echo "${alarm_list}" | grep " $ALARM_BASENAME$i " |
452 [ -n "$alarm_id" ] && $ceilometer alarm-delete "$alarm_id"
454 openstack $as_doctor_user subnet delete $NET_NAME
456 openstack $as_doctor_user network delete $NET_NAME
459 image_id=$(openstack image list | grep " $IMAGE_NAME " | awk '{print $2}')
461 #if an existing image was used, there's no need to remove it here
462 if [[ "$use_existing_image" == false ]] ; then
463 [ -n "$image_id" ] && openstack image delete "$image_id"
471 # NOTE: Temporal log printer.
472 for f in $(find . -name '*.log')
483 if $PYTHON_ENABLE; then
485 echo "executing tox..."
490 echo "Note: doctor/tests/run.sh has been executed."
491 git log --oneline -1 || true # ignore even you don't have git installed
495 source $TOP_DIR/functions-common
496 source $TOP_DIR/lib/installer
497 source $TOP_DIR/lib/inspector
501 echo "preparing VM image..."
505 echo "creating test user..."
508 echo "creating VM..."
512 echo "creating alarm..."
513 #TODO: change back to use, network problems depends on infra and installers
517 echo "starting doctor sample components..."
523 echo "injecting host failure..."
526 check_host_status "(DOWN|UNKNOWN)" 60
527 calculate_notification_time
528 unset_forced_down_hosts