# http://www.apache.org/licenses/LICENSE-2.0
##############################################################################
-[[ "${CI_DEBUG:-true}" == "true" ]] && set -x
+# Configuration
+
+[[ "${CI_DEBUG:-true}" == [Tt]rue ]] && set -x
IMAGE_URL=https://launchpad.net/cirros/trunk/0.3.0/+download/cirros-0.3.0-x86_64-disk.img
-IMAGE_NAME=cirros
+#if an existing image name is provided in the enviroment, use that one
+IMAGE_NAME=${IMAGE_NAME:-cirros}
IMAGE_FILE="${IMAGE_NAME}.img"
IMAGE_FORMAT=qcow2
VM_NAME=doctor_vm1
ALARM_NAME=doctor_alarm1
INSPECTOR_PORT=12345
CONSUMER_PORT=12346
-TEST_USER=demo
-TEST_PW=demo
-TEST_TENANT=demo
-TEST_ROLE=_member_
-
-SUPPORTED_INSTALLER_TYPES="apex local"
-INSTALLER_TYPE=${INSTALLER_TYPE:-apex}
-INSTALLER_IP=${INSTALLER_IP:-none}
-COMPUTE_HOST=${COMPUTE_HOST:-overcloud-novacompute-0}
-COMPUTE_IP=${COMPUTE_IP:-none}
-COMPUTE_USER=${COMPUTE_USER:-heat-admin}
-ssh_opts="-o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no"
+DOCTOR_USER=doctor
+DOCTOR_PW=doctor
+DOCTOR_PROJECT=doctor
+#TODO: change back to `_member_` when JIRA DOCTOR-55 is done
+DOCTOR_ROLE=admin
+PROFILER_TYPE=${PROFILER_TYPE:-none}
-if [[ ! "$SUPPORTED_INSTALLER_TYPES" =~ "$INSTALLER_TYPE" ]] ; then
- echo "ERROR: INSTALLER_TYPE=$INSTALLER_TYPE is not supported."
- exit 1
-fi
+TOP_DIR=$(cd $(dirname "$0") && pwd)
-prepare_compute_ssh() {
- ssh_opts_cpu="$ssh_opts"
+ssh_opts="-o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no"
+as_doctor_user="--os-username $DOCTOR_USER --os-password $DOCTOR_PW
+ --os-tenant-name $DOCTOR_PROJECT"
+
+
+# Functions
+
+get_compute_host_info() {
+ # get computer host info which VM boot in
+ COMPUTE_HOST=$(openstack $as_doctor_user server show $VM_NAME |
+ grep "OS-EXT-SRV-ATTR:host" | awk '{ print $4 }')
+ compute_host_in_undercloud=${COMPUTE_HOST%%.*}
+ die_if_not_set $LINENO COMPUTE_HOST "Failed to get compute hostname"
+
+ if is_installer apex; then
+ COMPUTE_USER=${COMPUTE_USER:-heat-admin}
+ COMPUTE_IP=$(sudo ssh $ssh_opts $INSTALLER_IP \
+ "source stackrc; \
+ nova show $compute_host_in_undercloud \
+ | awk '/ ctlplane network /{print \$5}'")
+ elif is_installer fuel; then
+ COMPUTE_USER=${COMPUTE_USER:-root}
+ node_id=$(echo $compute_host_in_undercloud | cut -d "-" -f 2)
+ COMPUTE_IP=$(sshpass -p r00tme ssh 2>/dev/null $ssh_opts root@${INSTALLER_IP} \
+ "fuel node|awk -F '|' -v id=$node_id '{if (\$1 == id) print \$5}' |xargs")
+ elif is_installer local; then
+ COMPUTE_USER=${COMPUTE_USER:-$(whoami)}
+ COMPUTE_IP=$(getent hosts "$COMPUTE_HOST" | awk '{ print $1 }')
+ fi
- if [[ "$INSTALLER_TYPE" == "apex" ]] ; then
- if [[ "$INSTALLER_IP" == "none" ]] ; then
- instack_mac=$(sudo virsh domiflist instack | awk '/default/{print $5}')
- INSTALLER_IP=$(/usr/sbin/arp -e | grep ${instack_mac} | awk '{print $1}')
- fi
+ die_if_not_set $LINENO COMPUTE_IP "Could not resolve $COMPUTE_HOST. Either manually set COMPUTE_IP or enable DNS resolution."
+ echo "COMPUTE_HOST=$COMPUTE_HOST"
+ echo "COMPUTE_IP=$COMPUTE_IP"
- if [[ "$COMPUTE_IP" == "none" ]] ; then
- COMPUTE_IP=$(sudo ssh $ssh_opts $INSTALLER_IP \
- "source stackrc; \
- nova show $COMPUTE_HOST \
- | awk '/ ctlplane network /{print \$5}'")
- fi
+ # verify connectivity to target compute host
+ ping -c 1 "$COMPUTE_IP"
+ if [[ $? -ne 0 ]] ; then
+ die $LINENO "Can not ping to computer host"
+ fi
- # get ssh key from installer node
- sudo scp $ssh_opts root@"$INSTALLER_IP":/home/stack/.ssh/id_rsa instack_key
- sudo chown $(whoami):$(whoami) instack_key
- chmod 400 instack_key
- ssh_opts_cpu+=" -i instack_key"
- elif [[ "$INSTALLER_TYPE" == "local" ]] ; then
- if [[ "$COMPUTE_IP" == "none" ]] ; then
- COMPUTE_IP=$(getent hosts "$COMPUTE_HOST" | awk '{ print $1 }')
- if [[ -z "$COMPUTE_IP" ]]; then
- echo "ERROR: Could not resolve $COMPUTE_HOST. Either manually set COMPUTE_IP or enable DNS resolution."
- exit 1
- fi
- fi
+ # verify ssh to target compute host
+ ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'exit'
+ if [[ $? -ne 0 ]] ; then
+ die $LINENO "Can not ssh to computer host"
+ fi
+}
- echo "INSTALLER_TYPE set to 'local'. Assuming SSH keys already exchanged with $COMPUTE_HOST"
+get_consumer_ip() {
+ local get_consumer_command="ip route get $COMPUTE_IP | awk '/ src /{print \$NF}'"
+ if is_installer apex; then
+ CONSUMER_IP=$(sudo ssh $ssh_opts root@$INSTALLER_IP \
+ "$get_consumer_command")
+ elif is_installer fuel; then
+ CONSUMER_IP=$(sudo sshpass -p r00tme ssh $ssh_opts root@${INSTALLER_IP} \
+ "$get_consumer_command")
+ elif is_installer local; then
+ CONSUMER_IP=`$get_consumer_command`
fi
+ echo "CONSUMER_IP=$CONSUMER_IP"
- # verify connectivity to target compute host
- ping -c 1 "$COMPUTE_IP"
+ die_if_not_set $LINENO CONSUMER_IP "Could not get CONSUMER_IP."
}
download_image() {
- [ -e "$IMAGE_FILE" ] && return 0
- wget "$IMAGE_URL" -o "$IMAGE_FILE"
+ #if a different name was provided for the image in the enviroment there's no need to download the image
+ use_existing_image=false
+ openstack image list | grep -q " $IMAGE_NAME " && use_existing_image=true
+
+ if [[ "$use_existing_image" == false ]] ; then
+ [ -e "$IMAGE_FILE" ] && return 0
+ wget "$IMAGE_URL" -o "$IMAGE_FILE"
+ fi
}
register_image() {
- glance image-list | grep -q " $IMAGE_NAME " && return 0
- glance image-create --name "$IMAGE_NAME" \
- --visibility public \
- --disk-format "$IMAGE_FORMAT" \
- --container-format bare \
- --file "$IMAGE_FILE"
+ openstack image list | grep -q " $IMAGE_NAME " && return 0
+ openstack image create "$IMAGE_NAME" \
+ --public \
+ --disk-format "$IMAGE_FORMAT" \
+ --container-format bare \
+ --file "$IMAGE_FILE"
}
create_test_user() {
- keystone user-list | grep -q "$TEST_USER" || {
- keystone user-create --name "$TEST_USER" --pass "$TEST_PW"
+ openstack project list | grep -q " $DOCTOR_PROJECT " || {
+ openstack project create "$DOCTOR_PROJECT"
}
- keystone tenant-list | grep -q "$TEST_TENANT" || {
- keystone tenant-create --name "$TEST_TENANT"
+ openstack user list | grep -q " $DOCTOR_USER " || {
+ openstack user create "$DOCTOR_USER" --password "$DOCTOR_PW" \
+ --project "$DOCTOR_PROJECT"
}
- keystone user-role-list --user "$TEST_USER" --tenant "$TEST_TENANT" \
- | grep -q "$TEST_ROLE" || {
- keystone user-role-add --user "$TEST_USER" --role "$TEST_ROLE" \
- --tenant "$TEST_TENANT"
+ openstack user role list "$DOCTOR_USER" --project "$DOCTOR_PROJECT" \
+ | grep -q " $DOCTOR_ROLE " || {
+ openstack role add "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
+ --project "$DOCTOR_PROJECT"
}
}
boot_vm() {
- nova list | grep -q " $VM_NAME " && return 0
- (
- # test VM done with test user, so can test non-admin
- export OS_USERNAME="$TEST_USER"
- export OS_PASSWORD="$TEST_PW"
- export OS_TENANT_NAME="$TEST_TENANT"
- nova boot --flavor "$VM_FLAVOR" \
- --image "$IMAGE_NAME" \
- "$VM_NAME"
- sleep 1
- )
-
+ # test VM done with test user, so can test non-admin
+ openstack $as_doctor_user server list | grep -q " $VM_NAME " && return 0
+ openstack $as_doctor_user server create --flavor "$VM_FLAVOR" \
+ --image "$IMAGE_NAME" \
+ "$VM_NAME"
+ sleep 1
}
create_alarm() {
- ceilometer alarm-list | grep -q " $ALARM_NAME " && return 0
- vm_id=$(nova list | grep " $VM_NAME " | awk '{print $2}')
- ceilometer alarm-event-create --name "$ALARM_NAME" \
+ # get vm_id as test user
+ ceilometer $as_doctor_user alarm-list | grep -q " $ALARM_NAME " && return 0
+ vm_id=$(openstack $as_doctor_user server list | grep " $VM_NAME " | awk '{print $2}')
+ # TODO(r-mibu): change notification endpoint from localhost to the consumer
+ # IP address (functest container).
+ ceilometer $as_doctor_user alarm-event-create --name "$ALARM_NAME" \
--alarm-action "http://localhost:$CONSUMER_PORT/failure" \
--description "VM failure" \
--enabled True \
start_monitor() {
pgrep -f "python monitor.py" && return 0
- sudo python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" \
+ sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
"http://127.0.0.1:$INSPECTOR_PORT/events" > monitor.log 2>&1 &
}
stop_monitor() {
pgrep -f "python monitor.py" || return 0
sudo kill $(pgrep -f "python monitor.py")
- cat monitor.log
-}
-
-start_inspector() {
- pgrep -f "python inspector.py" && return 0
- python inspector.py "$INSPECTOR_PORT" > inspector.log 2>&1 &
-}
-
-stop_inspector() {
- pgrep -f "python inspector.py" || return 0
- kill $(pgrep -f "python inspector.py")
- cat inspector.log
}
start_consumer() {
pgrep -f "python consumer.py" && return 0
python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 &
+
+ # NOTE(r-mibu): create tunnel to the controller nodes, so that we can
+ # avoid some network problems dpends on infra and installers.
+ # This tunnel will be terminated by stop_consumer() or after 10 mins passed.
+ if ! is_installer local; then
+ if is_installer apex; then
+ CONTROLLER_IPS=$(sudo ssh $ssh_opts $INSTALLER_IP \
+ "source stackrc; \
+ nova list | grep ' overcloud-controller-[0-9] ' \
+ | sed -e 's/^.*ctlplane=//' -e 's/ *|\$//'")
+ elif is_installer fuel; then
+ CONTROLLER_IPS=$(sshpass -p r00tme ssh 2>/dev/null $ssh_opts root@${INSTALLER_IP} \
+ "fuel node | grep controller | cut -d '|' -f 5|xargs")
+ fi
+
+ die_if_not_set $LINENO CONTROLLER_IPS "Could not get CONTROLLER_IPS."
+ for ip in $CONTROLLER_IPS
+ do
+ forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
+ tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
+ $tunnel_command > "ssh_tunnel.${ip}.log" 2>&1 < /dev/null &
+ done
+ fi
}
stop_consumer() {
pgrep -f "python consumer.py" || return 0
kill $(pgrep -f "python consumer.py")
- cat consumer.log
+
+ # NOTE(r-mibu): terminate tunnels to the controller nodes
+ if ! is_installer local; then
+ for ip in $CONTROLLER_IPS
+ do
+ forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
+ tunnel_command="sudo ssh $ssh_opts_cpu $COMPUTE_USER@$ip $forward_rule sleep 600"
+ kill $(pgrep -f "$tunnel_command")
+ done
+ fi
}
wait_for_vm_launch() {
echo "waiting for vm launch..."
- while true
+
+ count=0
+ while [[ ${count} -lt 60 ]]
do
- state=$(nova list | grep " $VM_NAME " | awk '{print $6}')
- [[ "$state" == "ACTIVE" ]] && return 0
+ state=$(openstack $as_doctor_user server list | grep " $VM_NAME " | awk '{print $6}')
+ if [[ "$state" == "ACTIVE" ]]; then
+ # NOTE(cgoncalves): sleeping for a bit to stabilize
+ # See python-openstackclient/functional/tests/compute/v2/test_server.py:wait_for_status
+ sleep 5
+ return 0
+ fi
+ if [[ "$state" == "ERROR" ]]; then
+ openstack $as_doctor_user server show $VM_NAME
+ die $LINENO "vm state is ERROR"
+ fi
+ count=$(($count+1))
sleep 1
done
+ die $LINENO "Time out while waiting for VM launch"
}
inject_failure() {
echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..."
cat > disable_network.sh << 'END_TXT'
#!/bin/bash -x
-dev=$(sudo ip route | awk '/^default/{print $5}')
+dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $7}')
sleep 1
sudo ip link set $dev down
+echo "doctor set host down at" $(date "+%s.%N")
sleep 180
sudo ip link set $dev up
sleep 1
END_TXT
+ sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh
chmod +x disable_network.sh
scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:"
ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &'
}
+profile_performance_poc() {
+ triggered=$(grep "^doctor set host down at" disable_network.log |\
+ sed -e "s/^.* at //")
+ vmdown=$(grep "doctor mark vm.* error at" inspector.log |tail -n 1 |\
+ sed -e "s/^.* at //")
+ hostdown=$(grep "doctor mark host.* down at" inspector.log |\
+ sed -e "s/^.* at //")
+
+ #calculate the relative interval to triggered(T00)
+ export DOCTOR_PROFILER_T00=0
+ export DOCTOR_PROFILER_T01=$(echo "($detected-$triggered)*1000/1" |bc)
+ export DOCTOR_PROFILER_T03=$(echo "($vmdown-$triggered)*1000/1" |bc)
+ export DOCTOR_PROFILER_T04=$(echo "($hostdown-$triggered)*1000/1" |bc)
+ export DOCTOR_PROFILER_T09=$(echo "($notified-$triggered)*1000/1" |bc)
+
+ python profiler-poc.py
+}
+
calculate_notification_time() {
- detected=$(grep "doctor monitor detected at" monitor.log | awk '{print $5}')
- notified=$(grep "doctor consumer notified at" consumer.log | awk '{print $5}')
+ if ! grep -q "doctor consumer notified at" consumer.log ; then
+ die $LINENO "Consumer hasn't received fault notification."
+ fi
+
+ #keep 'at' as the last keyword just before the value, and
+ #use regex to get value instead of the fixed column
+ detected=$(grep "doctor monitor detected at" monitor.log |\
+ sed -e "s/^.* at //")
+ notified=$(grep "doctor consumer notified at" consumer.log |\
+ sed -e "s/^.* at //")
+
+ if [[ "$PROFILER_TYPE" == "poc" ]]; then
+ profile_performance_poc
+ fi
+
echo "$notified $detected" | \
- awk '{d = $1 - $2; if (d < 1 && d > 0) print d " OK"; else print d " NG"}'
+ awk '{
+ d = $1 - $2;
+ if (d < 1 && d > 0) { print d " OK"; exit 0 }
+ else { print d " NG"; exit 1 }
+ }'
}
-check_host_status_down() {
- (
- # Switching to test user
- export OS_USERNAME="$TEST_USER"
- export OS_PASSWORD="$TEST_PW"
- export OS_TENANT_NAME="$TEST_TENANT"
-
- host_status_line=$(nova show $VM_NAME | grep "host_status")
- [[ $? -ne 0 ]] && {
- echo "ERROR: host_status not configured for owner in Nova policy.json"
- }
-
- host_status=$(echo $host_status_line | awk '{print $4}')
- [[ "$host_status" == "DOWN" ]] && {
- echo "$VM_NAME showing host_status: $host_status"
- }
- echo "ERROR: host_status not reported by: nova show $VM_NAME"
- )
+check_host_status() {
+ expected_state=$1
+
+ host_status_line=$(openstack $as_doctor_user --os-compute-api-version 2.16 \
+ server show $VM_NAME | grep "host_status")
+ host_status=$(echo $host_status_line | awk '{print $4}')
+ die_if_not_set $LINENO host_status "host_status not reported by: nova show $VM_NAME"
+ if [[ "$expected_state" =~ "$host_status" ]] ; then
+ echo "$VM_NAME showing host_status: $host_status"
+ else
+ die $LINENO "host_status:$host_status not equal to expected_state: $expected_state"
+ fi
+}
+
+unset_forced_down_hosts() {
+ for host in $(openstack compute service list --service nova-compute \
+ -f value -c Host -c State | sed -n -e '/down$/s/ *down$//p')
+ do
+ python ./nova_force_down.py $host --unset
+ done
+
+ echo "waiting disabled compute host back to be enabled..."
+ wait_until 'openstack compute service list --service nova-compute
+ -f value -c State | grep -q down' 240 5
}
cleanup() {
stop_inspector
stop_consumer
- python ./nova_force_down.py "$COMPUTE_HOST" --unset
- sleep 1
- nova delete "$VM_NAME"
+ unset_forced_down_hosts
+ # TODO: We need to make sure the target compute host is back to IP
+ # reachable. wait_ping() will be added by tojuvone .
+ sleep 110
+ scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
+
+ openstack $as_doctor_user server list | grep -q " $VM_NAME " && openstack $as_doctor_user server delete "$VM_NAME"
sleep 1
- alarm_id=$(ceilometer alarm-list | grep " $ALARM_NAME " | awk '{print $2}')
+ alarm_id=$(ceilometer $as_doctor_user alarm-list | grep " $ALARM_NAME " | awk '{print $2}')
sleep 1
- [ -n "$alarm_id" ] && ceilometer alarm-delete "$alarm_id"
+ [ -n "$alarm_id" ] && ceilometer $as_doctor_user alarm-delete "$alarm_id"
sleep 1
- image_id=$(glance image-list | grep " $IMAGE_NAME " | awk '{print $2}')
+
+ image_id=$(openstack image list | grep " $IMAGE_NAME " | awk '{print $2}')
sleep 1
- [ -n "$image_id" ] && glance image-delete "$image_id"
- keystone user-role-remove --user "$TEST_USER" --role "$TEST_ROLE" \
- --tenant "$TEST_TENANT"
- keystone tenant-remove --name "$TEST_TENANT"
- keystone user-delete "$TEST_USER"
+ #if an existing image was used, there's no need to remove it here
+ if [[ "$use_existing_image" == false ]] ; then
+ [ -n "$image_id" ] && openstack image delete "$image_id"
+ fi
+ openstack role remove "$DOCTOR_ROLE" --user "$DOCTOR_USER" \
+ --project "$DOCTOR_PROJECT"
+ openstack project delete "$DOCTOR_PROJECT"
+ openstack user delete "$DOCTOR_USER"
- #TODO: add host status check via nova admin api
- echo "waiting disabled compute host back to be enabled..."
- sleep 180
- ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" \
- "[ -e disable_network.log ] && cat disable_network.log"
+ cleanup_installer
+ cleanup_inspector
}
+# Main process
echo "Note: doctor/tests/run.sh has been executed."
-
-prepare_compute_ssh
+git log --oneline -1 || true # ignore even you don't have git installed
trap cleanup EXIT
+source $TOP_DIR/functions-common
+source $TOP_DIR/lib/installer
+source $TOP_DIR/lib/inspector
+
+setup_installer
+
echo "preparing VM image..."
download_image
register_image
-echo "starting doctor sample components..."
-start_monitor
-start_inspector
-start_consumer
-
echo "creating test user..."
create_test_user
-echo "creating VM and alarm..."
+echo "creating VM..."
boot_vm
-create_alarm
wait_for_vm_launch
+echo "get computer host info..."
+get_compute_host_info
+
+echo "creating alarm..."
+#TODO: change back to use, network problems depends on infra and installers
+#get_consumer_ip
+create_alarm
+
+echo "starting doctor sample components..."
+start_inspector
+start_monitor
+start_consumer
+
sleep 60
echo "injecting host failure..."
inject_failure
-sleep 10
+sleep 60
-check_host_status_down
+check_host_status "(DOWN|UNKNOWN)"
calculate_notification_time
echo "done"