#!/bin/bash -e ############################################################################## # Copyright (c) 2016 NEC Corporation and others. # # All rights reserved. This program and the accompanying materials # are made available under the terms of the Apache License, Version 2.0 # which accompanies this distribution, and is available at # http://www.apache.org/licenses/LICENSE-2.0 ############################################################################## # Configuration [[ "${CI_DEBUG:-true}" == [Tt]rue ]] && set -x IMAGE_URL=https://launchpad.net/cirros/trunk/0.3.0/+download/cirros-0.3.0-x86_64-disk.img #if an existing image name is provided in the enviroment, use that one IMAGE_NAME=${IMAGE_NAME:-cirros} IMAGE_FILE="${IMAGE_NAME}.img" IMAGE_FORMAT=qcow2 VM_NAME=doctor_vm1 VM_FLAVOR=m1.tiny ALARM_NAME=doctor_alarm1 INSPECTOR_PORT=12345 CONSUMER_PORT=12346 DOCTOR_USER=doctor DOCTOR_PW=doctor DOCTOR_PROJECT=doctor #TODO: change back to `_member_` when JIRA DOCTOR-55 is done DOCTOR_ROLE=admin PROFILER_TYPE=${PROFILER_TYPE:-none} TOP_DIR=$(cd $(dirname "$0") && pwd) 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" get_compute_ip_from_hostname $COMPUTE_HOST echo "COMPUTE_HOST=$COMPUTE_HOST" echo "COMPUTE_IP=$COMPUTE_IP" # verify connectivity to target compute host ping -c 1 "$COMPUTE_IP" if [[ $? -ne 0 ]] ; then die $LINENO "Can not ping to computer host" 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 } # 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 $COMPUTE_IP | awk '/ src /{print \$NF}'" if is_installer apex; then CONSUMER_IP=$(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" die_if_not_set $LINENO CONSUMER_IP "Could not get CONSUMER_IP." } download_image() { #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() { 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() { openstack project list | grep -q " $DOCTOR_PROJECT " || { openstack project create "$DOCTOR_PROJECT" } openstack user list | grep -q " $DOCTOR_USER " || { openstack user create "$DOCTOR_USER" --password "$DOCTOR_PW" \ --project "$DOCTOR_PROJECT" } 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() { # 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() { # 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:$CONSUMER_PORT/failure" \ --description "VM failure" \ --enabled True \ --repeat-actions False \ --severity "moderate" \ --event-type compute.instance.update \ -q "traits.state=string::error; traits.instance_id=string::$vm_id" } start_monitor() { pgrep -f "python monitor.py" && return 0 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 $(pgrep -f "python monitor.py") } start_consumer() { pgrep -f "python consumer.py" && return 0 python consumer.py "$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 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") # 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..." count=0 while [[ ${count} -lt 60 ]] do state=$(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 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 link 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@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &' # use host time to get rid of potential time sync deviation between nodes triggered=$(date "+%s.%N") } calculate_notification_time() { 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=$(grep "doctor consumer notified at" consumer.log |\ sed -e "s/^.* at //") echo "$notified $detected" | \ awk '{ d = $1 - $2; if (d < 1 && d > 0) { print d " OK"; exit 0 } else { print d " NG"; exit 1 } }' } 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$//p') do # TODO (r-mibu): make sample inspector use keystone v3 api OS_AUTH_URL=${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 } collect_logs() { 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@$COMPUTE_IP:disable_network.log" . # TODO(yujunz) collect other logs, e.g. nova, aodh } run_profiler() { if [[ "$PROFILER_TYPE" == "poc" ]]; then linkdown=$(grep "doctor set link down at " disable_network.log |\ sed -e "s/^.* at //") vmdown=$(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 //") # TODO(yujunz) check the actual delay to verify time sync status # expected ~1s delay from $trigger to $linkdown relative_start=${linkdown} export DOCTOR_PROFILER_T00=$(python -c \ "print(int(($linkdown-$relative_start)*1000))") export DOCTOR_PROFILER_T01=$(python -c \ "print(int(($detected-$relative_start)*1000))") export DOCTOR_PROFILER_T03=$(python -c \ "print(int(($vmdown-$relative_start)*1000))") export DOCTOR_PROFILER_T04=$(python -c \ "print(int(($hostdown-$relative_start)*1000))") export DOCTOR_PROFILER_T09=$(python -c \ "print(int(($notified-$relative_start)*1000))") python profiler-poc.py >doctor_profiler.log 2>&1 fi } cleanup() { set +e echo "cleanup..." stop_monitor stop_inspector stop_consumer 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 alarm_id=$(ceilometer $as_doctor_user alarm-list | grep " $ALARM_NAME " | awk '{print $2}') sleep 1 [ -n "$alarm_id" ] && ceilometer $as_doctor_user alarm-delete "$alarm_id" sleep 1 image_id=$(openstack image list | grep " $IMAGE_NAME " | awk '{print $2}') sleep 1 #if an existing image was used, there's no need to remove it here if [[ "$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" 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 setup_installer echo "preparing VM image..." download_image register_image echo "creating test user..." create_test_user echo "creating VM..." boot_vm 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 60 check_host_status "(DOWN|UNKNOWN)" calculate_notification_time collect_logs run_profiler echo "done"