# http://www.apache.org/licenses/LICENSE-2.0
##############################################################################
+# Configuration
+
[[ "${CI_DEBUG:-true}" == [Tt]rue ]] && set -x
IMAGE_URL=https://launchpad.net/cirros/trunk/0.3.0/+download/cirros-0.3.0-x86_64-disk.img
DOCTOR_PROJECT=doctor
#TODO: change back to `_member_` when JIRA DOCTOR-55 is done
DOCTOR_ROLE=admin
+PROFILER_TYPE=${PROFILER_TYPE:-none}
-SUPPORTED_INSTALLER_TYPES="apex fuel local"
-INSTALLER_TYPE=${INSTALLER_TYPE:-local}
-INSTALLER_IP=${INSTALLER_IP:-none}
-
-SUPPORTED_INSPECTOR_TYPES="sample congress"
-INSPECTOR_TYPE=${INSPECTOR_TYPE:-sample}
TOP_DIR=$(cd $(dirname "$0") && pwd)
-ssh_opts="-o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no"
as_doctor_user="--os-username $DOCTOR_USER --os-password $DOCTOR_PW
--os-tenant-name $DOCTOR_PROJECT"
-if [[ ! "$SUPPORTED_INSTALLER_TYPES" =~ "$INSTALLER_TYPE" ]] ; then
- echo "ERROR: INSTALLER_TYPE=$INSTALLER_TYPE is not supported."
- exit 1
-fi
-
-if [[ ! "$SUPPORTED_INSPECTOR_TYPES" =~ "$INSPECTOR_TYPE" ]] ; then
- echo "ERROR: INSPECTOR_TYPE=$INSPECTOR_TYPE is not supported."
- exit 1
-fi
-
-get_installer_ip() {
- if [[ "$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 $1}')
- fi
- elif [[ "$INSTALLER_TYPE" == "fuel" ]] ; then
- if [[ "$INSTALLER_IP" == "none" ]] ; then
- instack_mac=$(sudo virsh domiflist fuel-opnfv | awk '/pxebr/{print $5}')
- INSTALLER_IP=$(/usr/sbin/arp -e | grep ${instack_mac} | awk '{print $1}')
- fi
- fi
-
- if [[ "$INSTALLER_TYPE" != "local" ]] ; then
- die_if_not_set $LINENO INSTALLER_IP "No installer IP"
- fi
-}
-
-prepare_ssh_to_cloud() {
- ssh_opts_cpu="$ssh_opts"
-
- # get ssh key from installer node
- if [[ "$INSTALLER_TYPE" == "apex" ]] ; then
- 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" == "fuel" ]] ; then
- sshpass -p r00tme scp $ssh_opts root@${INSTALLER_IP}:.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
- echo "INSTALLER_TYPE set to 'local'. Assuming SSH keys already exchanged with $COMPUTE_HOST"
- fi
-}
-
-prepare_test_env() {
- #TODO delete it when fuel support the configuration
- if [[ "$INSTALLER_TYPE" == "fuel" ]] ; then
- echo "modify the configuration..."
- cat > set_conf.sh << 'END_TXT'
-#!/bin/bash
-if [ -e /etc/ceilometer/event_pipeline.yaml ]; then
- if ! grep -q '^ *- notifier://?topic=alarm.all$' /etc/ceilometer/event_pipeline.yaml; then
- sed -i 's|- notifier://|- notifier://?topic=alarm.all|' /etc/ceilometer/event_pipeline.yaml
- echo "modify the ceilometer config"
- service ceilometer-agent-notification restart
- fi
-else
- echo "ceilometer event_pipeline.yaml file does not exist"
- exit 1
-fi
-if [ -e /etc/nova/nova.conf ]; then
- if ! grep -q '^notification_driver=messaging$' /etc/nova/nova.conf; then
- sed -i -r 's/notification_driver=/notification_driver=messaging/g' /etc/nova/nova.conf
- echo "modify nova config"
- service nova-api restart
- fi
-else
- echo "nova.conf file does not exist"
- exit 1
-fi
-exit 0
-END_TXT
- chmod +x set_conf.sh
- CONTROLLER_IP=$(sshpass -p r00tme ssh 2>/dev/null $ssh_opts root@${INSTALLER_IP} \
- "fuel node | grep controller | cut -d '|' -f 5|xargs")
- for node in $CONTROLLER_IP;do
- scp $ssh_opts_cpu set_conf.sh "root@$node:"
- ssh $ssh_opts_cpu "root@$node" './set_conf.sh > set_conf.log 2>&1 &'
- sleep 1
- scp $ssh_opts_cpu "root@$node:set_conf.log" set_conf_$node.log
- done
-
- if grep -q "modify the ceilometer config" set_conf_*.log ; then
- NEED_TO_RESTORE_CEILOMETER=true
- fi
- if grep -q "modify nova config" set_conf_*.log ; then
- NEED_TO_RESTORE_NOVA=true
- fi
-
- echo "waiting service restart..."
- sleep 60
- fi
-}
-restore_test_env() {
- #TODO delete it when fuel support the configuration
- if [[ "$INSTALLER_TYPE" == "fuel" ]] ; then
- if ! ($NEED_TO_RESTORE_CEILOMETER || $NEED_TO_RESTORE_NOVA) ; then
- echo "Don't need to restore config"
- exit 0
- fi
-
- echo "restore the configuration..."
- cat > restore_conf.sh << 'END_TXT'
-#!/bin/bash
-if @NEED_TO_RESTORE_CEILOMETER@ ; then
- if [ -e /etc/ceilometer/event_pipeline.yaml ]; then
- if grep -q '^ *- notifier://?topic=alarm.all$' /etc/ceilometer/event_pipeline.yaml; then
- sed -i 's|- notifier://?topic=alarm.all|- notifier://|' /etc/ceilometer/event_pipeline.yaml
- service ceilometer-agent-notification restart
- fi
- else
- echo "ceilometer event_pipeline.yaml file does not exist"
- exit 1
- fi
-fi
-if @NEED_TO_RESTORE_NOVA@ ; then
- if [ -e /etc/nova/nova.conf ]; then
- if grep -q '^notification_driver=messaging$' /etc/nova/nova.conf; then
- sed -i -r 's/notification_driver=messaging/notification_driver=/g' /etc/nova/nova.conf
- service nova-api restart
- fi
- else
- echo "nova.conf file does not exist"
- exit 1
- fi
-fi
-exit 0
-END_TXT
- sed -i -e "s/@NEED_TO_RESTORE_CEILOMETER@/$NEED_TO_RESTORE_CEILOMETER/" restore_conf.sh
- sed -i -e "s/@NEED_TO_RESTORE_NOVA@/$NEED_TO_RESTORE_NOVA/" restore_conf.sh
- chmod +x restore_conf.sh
- for node in $CONTROLLER_IP;do
- scp $ssh_opts_cpu restore_conf.sh "root@$node:"
- ssh $ssh_opts_cpu "root@$node" './restore_conf.sh > restore_conf.log 2>&1 &'
- done
-
- echo "waiting service restart..."
- sleep 60
- fi
-}
+# Functions
get_compute_host_info() {
# get computer host info which VM boot in
compute_host_in_undercloud=${COMPUTE_HOST%%.*}
die_if_not_set $LINENO COMPUTE_HOST "Failed to get compute hostname"
- if [[ "$INSTALLER_TYPE" == "apex" ]] ; then
- COMPUTE_USER=${COMPUTE_USER:-heat-admin}
- COMPUTE_IP=$(sudo ssh $ssh_opts $INSTALLER_IP \
- "source stackrc; \
- nova show $compute_host_in_undercloud \
- | awk '/ ctlplane network /{print \$5}'")
- elif [[ "$INSTALLER_TYPE" == "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=$node_id '{if (\$1 == id) print \$5}' |xargs")
- elif [[ "$INSTALLER_TYPE" == "local" ]] ; then
- COMPUTE_USER=${COMPUTE_USER:-$(whoami)}
- COMPUTE_IP=$(getent hosts "$COMPUTE_HOST" | awk '{ print $1 }')
- fi
+ get_compute_ip_from_hostname $COMPUTE_HOST
- 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"
fi
}
-get_consumer_ip() {
+# TODO(r-mibu): update this function to support consumer instance
+# and migrate this function into installer lib
+get_consumer_ip___to_be_removed() {
local get_consumer_command="ip route get $COMPUTE_IP | awk '/ src /{print \$NF}'"
- if [[ "$INSTALLER_TYPE" == "apex" ]] ; then
+ if is_installer apex; then
CONSUMER_IP=$(sudo ssh $ssh_opts root@$INSTALLER_IP \
"$get_consumer_command")
- elif [[ "$INSTALLER_TYPE" == "fuel" ]] ; then
+ elif is_installer fuel; then
CONSUMER_IP=$(sudo sshpass -p r00tme ssh $ssh_opts root@${INSTALLER_IP} \
"$get_consumer_command")
- elif [[ "$INSTALLER_TYPE" == "local" ]] ; then
+ elif is_installer local; then
CONSUMER_IP=`$get_consumer_command`
fi
echo "CONSUMER_IP=$CONSUMER_IP"
-q "traits.state=string::error; traits.instance_id=string::$vm_id"
}
-print_log() {
- log_file=$1
- echo "$log_file:"
- sed -e 's/^/ /' "$log_file"
-}
-
start_monitor() {
pgrep -f "python monitor.py" && return 0
sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
stop_monitor() {
pgrep -f "python monitor.py" || return 0
sudo kill $(pgrep -f "python monitor.py")
- print_log monitor.log
-}
-
-congress_add_rule() {
- name=$1
- policy=$2
- rule=$3
-
- if ! openstack congress policy rule list $policy | grep -q -e "// Name: $name$" ; then
- openstack congress policy rule create --name $name $policy "$rule"
- fi
-}
-
-congress_del_rule() {
- name=$1
- policy=$2
-
- if openstack congress policy rule list $policy | grep -q -e "^// Name: $name$" ; then
- openstack congress policy rule delete $policy $name
- fi
-}
-
-congress_setup_rules() {
- congress_add_rule host_down classification \
- 'host_down(host) :-
- doctor:events(hostname=host, type="compute.host.down", status="down")'
-
- congress_add_rule active_instance_in_host classification \
- 'active_instance_in_host(vmid, host) :-
- nova:servers(id=vmid, host_name=host, status="ACTIVE")'
-
- congress_add_rule host_force_down classification \
- 'execute[nova:services.force_down(host, "nova-compute", "True")] :-
- host_down(host)'
-
- congress_add_rule error_vm_states classification \
- 'execute[nova:servers.reset_state(vmid, "error")] :-
- host_down(host),
- active_instance_in_host(vmid, host)'
-}
-
-start_inspector() {
- if [[ "$INSPECTOR_TYPE" == "sample" ]] ; then
- pgrep -f "python inspector.py" && return 0
- python inspector.py "$INSPECTOR_PORT" > inspector.log 2>&1 &
- elif [[ "$INSPECTOR_TYPE" == "congress" ]] ; then
- nova_api_min_version="2.11"
- nova_api_version=$(openstack congress datasource list | \
- grep nova | grep -Po "(?<='api_version': ')[^']*")
- if ! is_set nova_api_version; then
- nova_api_version="2.0"
- fi
- if [[ "$nova_api_version" < "$nova_api_min_version" ]]; then
- die $LINENO "Congress Nova datasource API version < $nova_api_min_version ($nova_api_version)"
- fi
- openstack congress driver list | grep -q " doctor "
- openstack congress datasource list | grep -q " doctor " || {
- openstack congress datasource create doctor doctor
- }
- congress_setup_rules
- fi
-}
-
-stop_inspector() {
- if [[ "$INSPECTOR_TYPE" == "sample" ]] ; then
- pgrep -f "python inspector.py" || return 0
- kill $(pgrep -f "python inspector.py")
- print_log inspector.log
- elif [[ "$INSPECTOR_TYPE" == "congress" ]] ; then
- congress_del_rule host_force_down classification
- congress_del_rule error_vm_states classification
- congress_del_rule active_instance_in_host classification
- congress_del_rule host_down classification
- fi
}
start_consumer() {
# 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 [[ "$INSTALLER_TYPE" != "local" ]] ; then
- if [[ "$INSTALLER_TYPE" == "apex" ]] ; then
- CONTROLLER_IPS=$(sudo ssh $ssh_opts $INSTALLER_IP \
- "source stackrc; \
- nova list | grep ' overcloud-controller-[0-9] ' \
- | sed -e 's/^.*ctlplane=//' -e 's/ *|\$//'")
- elif [[ "$INSTALLER_TYPE" == "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."
+ if ! is_installer local; then
for ip in $CONTROLLER_IPS
do
forward_rule="-R $CONSUMER_PORT:localhost:$CONSUMER_PORT"
stop_consumer() {
pgrep -f "python consumer.py" || return 0
kill $(pgrep -f "python consumer.py")
- print_log consumer.log
# NOTE(r-mibu): terminate tunnels to the controller nodes
- if [[ "$INSTALLER_TYPE" != "local" ]] ; then
+ 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")
- print_log "ssh_tunnel.${ip}.log"
done
fi
}
return 0
fi
if [[ "$state" == "ERROR" ]]; then
+ openstack $as_doctor_user server show $VM_NAME
die $LINENO "vm state is ERROR"
fi
count=$(($count+1))
cat > disable_network.sh << 'END_TXT'
#!/bin/bash -x
dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $7}')
+[[ -n "$dev" ]] || dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $5}')
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
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 $10}')
- notified=$(grep "doctor consumer notified at" consumer.log | awk '{print $10}')
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;
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$//p')
+ do
+ # TODO (r-mibu): make sample inspector use keystone v3 api
+ OS_AUTH_URL=${OS_AUTH_URL/v3/v2.0} \
+ python ./nova_force_down.py $host --unset
+ done
+
+ echo "waiting disabled compute host back to be enabled..."
+ wait_until 'openstack compute service list --service nova-compute
+ -f value -c State | grep -q down' 240 5
+}
+
cleanup() {
set +e
echo "cleanup..."
stop_inspector
stop_consumer
- echo "waiting disabled compute host back to be enabled..."
- python ./nova_force_down.py "$COMPUTE_HOST" --unset
- sleep 240
- check_host_status "UP"
- scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
- print_log disable_network.log
+ 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
+ if is_set COMPUTE_IP; then
+ scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:disable_network.log" .
+ fi
openstack $as_doctor_user server list | grep -q " $VM_NAME " && openstack $as_doctor_user server delete "$VM_NAME"
sleep 1
openstack project delete "$DOCTOR_PROJECT"
openstack user delete "$DOCTOR_USER"
- restore_test_env
+ cleanup_installer
+ cleanup_inspector
}
+# Main process
echo "Note: doctor/tests/run.sh has been executed."
+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
-echo "preparing test env..."
-get_installer_ip
-prepare_ssh_to_cloud
-prepare_test_env
+setup_installer
echo "preparing VM image..."
download_image
Code Review / doctor.git / blobdiff