test/vhost: add live migration test case 2

Change-Id: I114c11d9f7c1e9a5a8c0a989541a41b8747cc125
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-on: https://review.gerrithub.io/398753
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
This commit is contained in:
Pawel Wodkowski 2018-02-01 22:14:58 +01:00 committed by Daniel Verkamp
parent 90922c60df
commit 6ebfbf7351
5 changed files with 251 additions and 9 deletions

View File

@ -112,6 +112,7 @@ function spdk_vhost_run()
local param local param
local vhost_num=0 local vhost_num=0
local vhost_conf_path="" local vhost_conf_path=""
local memory=1024
for param in "$@"; do for param in "$@"; do
case $param in case $param in
@ -120,6 +121,7 @@ function spdk_vhost_run()
assert_number "$vhost_num" assert_number "$vhost_num"
;; ;;
--conf-path=*) local vhost_conf_path="${param#*=}" ;; --conf-path=*) local vhost_conf_path="${param#*=}" ;;
--memory=*) local memory=${param#*=} ;;
*) *)
error "Invalid parameter '$param'" error "Invalid parameter '$param'"
return 1 return 1
@ -127,7 +129,6 @@ function spdk_vhost_run()
esac esac
done done
local vhost_dir="$(get_vhost_dir $vhost_num)" local vhost_dir="$(get_vhost_dir $vhost_num)"
if [[ -z "$vhost_conf_path" ]]; then if [[ -z "$vhost_conf_path" ]]; then
error "Missing mandatory parameter '--conf-path'" error "Missing mandatory parameter '--conf-path'"
@ -163,13 +164,14 @@ function spdk_vhost_run()
cp $vhost_conf_template $vhost_conf_file cp $vhost_conf_template $vhost_conf_file
$SPDK_BUILD_DIR/scripts/gen_nvme.sh >> $vhost_conf_file $SPDK_BUILD_DIR/scripts/gen_nvme.sh >> $vhost_conf_file
local cmd="$vhost_app -m $reactor_mask -p $master_core -c $vhost_conf_file -r $vhost_dir/rpc.sock" local cmd="$vhost_app -m $reactor_mask -p $master_core -c $vhost_conf_file -s $memory -r $vhost_dir/rpc.sock"
notice "Loging to: $vhost_log_file" notice "Loging to: $vhost_log_file"
notice "Config file: $vhost_conf_file" notice "Config file: $vhost_conf_file"
notice "Socket: $vhost_socket" notice "Socket: $vhost_socket"
notice "Command: $cmd" notice "Command: $cmd"
timing_enter vhost_start
cd $vhost_dir; $cmd & cd $vhost_dir; $cmd &
vhost_pid=$! vhost_pid=$!
echo $vhost_pid > $vhost_pid_file echo $vhost_pid > $vhost_pid_file
@ -177,12 +179,14 @@ function spdk_vhost_run()
notice "waiting for app to run..." notice "waiting for app to run..."
waitforlisten "$vhost_pid" "$vhost_dir/rpc.sock" waitforlisten "$vhost_pid" "$vhost_dir/rpc.sock"
notice "vhost started - pid=$vhost_pid" notice "vhost started - pid=$vhost_pid"
timing_exit vhost_start
rm $vhost_conf_file rm $vhost_conf_file
} }
function spdk_vhost_kill() function spdk_vhost_kill()
{ {
local rc=0
local vhost_num=0 local vhost_num=0
if [[ ! -z "$1" ]]; then if [[ ! -z "$1" ]]; then
vhost_num=$1 vhost_num=$1
@ -196,6 +200,7 @@ function spdk_vhost_kill()
return 0 return 0
fi fi
timing_enter vhost_kill
local vhost_pid="$(cat $vhost_pid_file)" local vhost_pid="$(cat $vhost_pid_file)"
notice "killing vhost (PID $vhost_pid) app" notice "killing vhost (PID $vhost_pid) app"
@ -213,19 +218,24 @@ function spdk_vhost_kill()
error "ERROR: vhost was NOT killed - sending SIGABRT" error "ERROR: vhost was NOT killed - sending SIGABRT"
/bin/kill -ABRT $vhost_pid /bin/kill -ABRT $vhost_pid
rm $vhost_pid_file rm $vhost_pid_file
return 1 rc=1
else
#check vhost return code, activate trap on error
wait $vhost_pid
fi fi
#check vhost return code, activate trap on error
wait $vhost_pid
elif /bin/kill -0 $vhost_pid; then elif /bin/kill -0 $vhost_pid; then
error "vhost NOT killed - you need to kill it manually" error "vhost NOT killed - you need to kill it manually"
return 1 rc=1
else else
notice "vhost was no running" notice "vhost was no running"
fi fi
rm $vhost_pid_file timing_exit vhost_kill
if [[ $rc == 0 ]]; then
rm $vhost_pid_file
fi
return $rc
} }
### ###

View File

@ -0,0 +1,20 @@
[global]
blocksize_range=4k-512k
iodepth=128
ioengine=libaio
filename=
group_reporting
thread
numjobs=1
direct=1
do_verify=1
verify=md5
verify_fatal=1
verify_dump=1
verify_backlog=8
[randwrite]
rw=randwrite
runtime=15
time_based
stonewall

View File

@ -0,0 +1,207 @@
source $SPDK_BUILD_DIR/test/nvmf/common.sh
function migration_tc2_cleanup_nvmf_tgt()
{
local i
if [[ ! -r "$nvmf_dir/nvmf_tgt.pid" ]]; then
warning "Pid file '$nvmf_dir/nvmf_tgt.pid' does not exist. "
return
fi
if [[ ! -z "$1" ]]; then
trap 'error_exit "${FUNCNAME}" "${LINENO}"' INT ERR EXIT
pkill --signal $1 -F $nvmf_dir/nvmf_tgt.pid || true
sleep 5
if ! pkill -F $nvmf_dir/nvmf_tgt.pid; then
fail "failed to kill nvmf_tgt app"
fi
else
pkill --signal SIGTERM -F $nvmf_dir/nvmf_tgt.pid || true
for (( i=0; i<20; i++ )); do
if ! pkill --signal 0 -F $nvmf_dir/nvmf_tgt.pid; then
break
fi
sleep 0.5
done
if pkill --signal 0 -F $nvmf_dir/nvmf_tgt.pid; then
error "nvmf_tgt failed to shutdown"
fi
fi
rm $nvmf_dir/nvmf_tgt.pid
unset -v nvmf_dir rpc_nvmf
}
function migration_tc2_cleanup_vhost_config()
{
timing_enter migration_tc2_cleanup_vhost_config
trap 'migration_tc2_cleanup_nvmf_tgt SIGKILL; error_exit "${FUNCNAME}" "${LINENO}"' INT ERR EXIT
notice "Shutting down all VMs"
vm_shutdown_all
notice "Removing vhost devices & controllers via RPC ..."
# Delete bdev first to remove all LUNs and SCSI targets
$rpc_0 delete_bdev Nvme0n1
$rpc_0 remove_vhost_controller $incoming_vm_ctrlr
$rpc_1 delete_bdev Nvme0n1
$rpc_1 remove_vhost_controller $target_vm_ctrlr
notice "killing vhost app"
spdk_vhost_kill 0
spdk_vhost_kill 1
unset -v incoming_vm target_vm incoming_vm_ctrlr target_vm_ctrlr
unset -v rpc_0 rpc_1
trap 'error_exit "${FUNCNAME}" "${LINENO}"' INT ERR EXIT
migration_tc2_cleanup_nvmf_tgt
timing_exit migration_tc2_cleanup_vhost_config
}
function migration_tc2_configure_vhost()
{
timing_enter migration_tc2_configure_vhost
# Those are global intentionaly - they will be unset in cleanup handler
nvmf_dir="$TEST_DIR/nvmf_tgt"
incoming_vm=1
target_vm=2
incoming_vm_ctrlr=naa.VhostScsi0.$incoming_vm
target_vm_ctrlr=naa.VhostScsi0.$target_vm
rpc_nvmf="python $SPDK_BUILD_DIR/scripts/rpc.py -s $nvmf_dir/rpc.sock"
rpc_0="python $SPDK_BUILD_DIR/scripts/rpc.py -s $(get_vhost_dir 0)/rpc.sock"
rpc_1="python $SPDK_BUILD_DIR/scripts/rpc.py -s $(get_vhost_dir 1)/rpc.sock"
# Default cleanup/error handlers will not shutdown nvmf_tgt app so setup it
# here to teardown in cleanup function
trap 'migration_tc2_error_cleanup; error_exit "${FUNCNAME}" "${LINENO}"' INT ERR EXIT
# Run nvmf_tgt and two vhost instances:
# nvmf_tgt uses core id 2 (-m 0x4)
# First uses core id 0 (vhost_0_reactor_mask=0x1)
# Second uses core id 1 (vhost_1_reactor_mask=0x2)
# This force to use VM 1 and 2.
timing_enter start_nvmf_tgt
notice "Running nvmf_tgt..."
mkdir -p $nvmf_dir
rm -f $nvmf_dir/*
cp $SPDK_BUILD_DIR/test/nvmf/nvmf.conf $nvmf_dir/nvmf.conf
$SPDK_BUILD_DIR/scripts/gen_nvme.sh >> $nvmf_dir/nvmf.conf
$SPDK_BUILD_DIR/app/nvmf_tgt/nvmf_tgt -s 512 -m 0x4 -c $nvmf_dir/nvmf.conf -r $nvmf_dir/rpc.sock &
local nvmf_tgt_pid=$!
echo $nvmf_tgt_pid > $nvmf_dir/nvmf_tgt.pid
waitforlisten "$nvmf_tgt_pid" "$nvmf_dir/rpc.sock"
timing_exit start_nvmf_tgt
spdk_vhost_run --conf-path=$BASE_DIR --memory=512 --vhost-num=0
# Those are global intentionaly
vhost_1_reactor_mask=0x2
vhost_1_master_core=1
spdk_vhost_run --conf-path=$BASE_DIR --memory=512 --vhost-num=1
local rdma_ip_list=$(get_available_rdma_ips)
local nvmf_target_ip=$(echo "$rdma_ip_list" | head -n 1)
if [[ -z "$nvmf_target_ip" ]]; then
fail "no NIC for nvmf target"
fi
notice "Configuring nvmf_tgt, vhost devices & controllers via RPC ..."
# Construct shared bdevs and controllers
$rpc_nvmf construct_nvmf_subsystem nqn.2016-06.io.spdk:cnode1 \
"trtype:RDMA traddr:$nvmf_target_ip trsvcid:4420" "" -a -s SPDK00000000000001 -n Nvme0n1
$rpc_0 construct_nvme_bdev -b Nvme0 -t rdma -f ipv4 -a $nvmf_target_ip -s 4420 -n "nqn.2016-06.io.spdk:cnode1"
$rpc_0 construct_vhost_scsi_controller $incoming_vm_ctrlr
$rpc_0 add_vhost_scsi_lun $incoming_vm_ctrlr 0 Nvme0n1
$rpc_1 construct_nvme_bdev -b Nvme0 -t rdma -f ipv4 -a $nvmf_target_ip -s 4420 -n "nqn.2016-06.io.spdk:cnode1"
$rpc_1 construct_vhost_scsi_controller $target_vm_ctrlr
$rpc_1 add_vhost_scsi_lun $target_vm_ctrlr 0 Nvme0n1
notice "Setting up VMs"
vm_setup --os="$os_image" --force=$incoming_vm --disk-type=spdk_vhost_scsi --disks=VhostScsi0 \
--migrate-to=$target_vm --memory=1024 --vhost-num=0
vm_setup --force=$target_vm --disk-type=spdk_vhost_scsi --disks=VhostScsi0 --incoming=$incoming_vm --memory=1024 \
--vhost-num=1
# Run everything
vm_run $incoming_vm $target_vm
# Wait only for incoming VM, as target is waiting for migration
vm_wait_for_boot 600 $incoming_vm
notice "Configuration done"
timing_exit migration_tc2_configure_vhost
}
function migration_tc2_error_cleanup()
{
trap - SIGINT ERR EXIT
set -x
vm_kill_all
migration_tc2_cleanup_vhost_config
notice "Migration TC2 FAILED"
}
function migration_tc2()
{
# Use 2 VMs:
# incoming VM - the one we want to migrate
# targe VM - the one which will accept migration
local job_file="$BASE_DIR/migration-tc2.job"
migration_tc2_configure_vhost
# Run fio before migration
notice "Starting FIO"
vm_check_scsi_location $incoming_vm
run_fio $fio_bin --job-file="$job_file" --local --vm="${incoming_vm}$(printf ':/dev/%s' $SCSI_DISK)"
# Wait a while to let the FIO time to issue some IO
sleep 5
# Check if fio is still running before migration
if ! is_fio_running $incoming_vm; then
vm_ssh $incoming_vm "cat /root/$(basename ${job_file}).out"
error "FIO is not running before migration: process crashed or finished too early"
fi
vm_migrate $incoming_vm
sleep 3
# Check if fio is still running after migration
if ! is_fio_running $target_vm; then
vm_ssh $target_vm "cat /root/$(basename ${job_file}).out"
error "FIO is not running after migration: process crashed or finished too early"
fi
notice "Waiting for fio to finish"
local timeout=40
while is_fio_running $target_vm; do
sleep 1
echo -n "."
if (( timeout-- == 0 )); then
error "timeout while waiting for FIO!"
fi
done
notice "Fio result is:"
vm_ssh $target_vm "cat /root/$(basename ${job_file}).out"
migration_tc2_cleanup_vhost_config
notice "Migration TC2 SUCCESS"
}
migration_tc2

View File

@ -76,6 +76,7 @@ function vm_migrate()
fail "source VM $1 or destination VM is not properly configured for live migration" fail "source VM $1 or destination VM is not properly configured for live migration"
fi fi
timing_enter vm_migrate
notice "Migrating VM $1 to VM "$(basename $target_vm_dir) notice "Migrating VM $1 to VM "$(basename $target_vm_dir)
echo -e \ echo -e \
"migrate_set_speed 1g\n" \ "migrate_set_speed 1g\n" \
@ -95,6 +96,7 @@ function vm_migrate()
fi fi
notice "Migration complete" notice "Migration complete"
timing_exit vm_migrate
} }
function is_fio_running() function is_fio_running()
@ -120,7 +122,10 @@ for test_case in ${test_cases//,/ }; do
notice "===============================" notice "==============================="
notice "Running Migration test case ${test_case}" notice "Running Migration test case ${test_case}"
notice "===============================" notice "==============================="
timing_enter migration-tc${test_case}
source $BASE_DIR/migration-tc${test_case}.sh source $BASE_DIR/migration-tc${test_case}.sh
timing_exit migration-tc${test_case}
done done
notice "Migration Test SUCCESS" notice "Migration Test SUCCESS"

View File

@ -85,7 +85,7 @@ case $1 in
-m|--migration) -m|--migration)
echo 'Running migration suite...' echo 'Running migration suite...'
$WORKDIR/migration/migration.sh -x \ $WORKDIR/migration/migration.sh -x \
--fio-bin=$FIO_BIN --os=$VM_IMAGE --test-cases=1 --fio-bin=$FIO_BIN --os=$VM_IMAGE --test-cases=1,2
;; ;;
-i|--integrity) -i|--integrity)
echo 'Running SCSI integrity suite...' echo 'Running SCSI integrity suite...'